Abstract
Introduction:
Cirrhosis remains a major burden on the healthcare system despite substantial advances in therapy and care. Studies simultaneously examining mortality, readmission, and cost of care are not available. Here, we hypothesized that improved patient care in the last decade might have led to improved outcomes and reduced costs in patients with cirrhosis.
Methods:
We identified compensated cirrhosis (CC) and decompensated cirrhosis (DC) patients using carefully chosen ICD-9/ICD-10 codes from the Nationwide Readmission Database (NRD) (years 2010 to 2016). We evaluated trends of 30-day all-cause mortality, 30-day readmission, and inflation-adjusted index hospitalization and readmission costs. Factors associated with mortality and readmission were identified using regression analyses.
Results:
3,374,038 patients with cirrhosis were identified, of whom nearly 50% had a decompensating event on initial admission. The 30-day inpatient mortality rate for both CC and DC patients decreased from 2010 to 2016. The 30-day readmission rate remained stable for DC and declined for CC. Over the study period, 30-day readmission costs increased for DC and remained unchanged for CC. The median cost for index hospitalization remained nearly unchanged, but the cost of readmission increased for both CC and DC groups. Gastrointestinal diseases and infections were the leading cause of readmission in CC and DC patient groups.
Conclusion:
Inpatient mortality has decreased for CC and DC patients . Readmission has declined for CC patients and remained stable for DC patients. However, the economic burden of cirrhosis is rising.
Keywords: Outcomes Research, 30 Day Readmissions, Portal Hypertension, Patient care, Liver diseases
Introduction
Cirrhosis, the result of ongoing fibrogenesis in the liver, is the final common result of the majority of chronic liver diseases[1]. Cirrhosis and its complications are implicated as the cause of over 40,000 deaths annually and are associated with significant morbidity, mortality, and healthcare costs based on US population studies[1, 2]. US studies estimate the healthcare costs associated with advanced liver disease ranges from $2.5 to $10.6 billion annually[3–5]. Clinically, cirrhosis is classified into two main stages: compensated cirrhosis (CC) and decompensated cirrhosis (DC)[6]. The transition from compensated to decompensated cirrhosis happens at the yearly rate of 5-7%[6]. This transition represents a critical event since median survival drops from 12 years in the compensated stage to 1.5 years in patients with decompensated cirrhosis[2, 6].
Due to the chronic nature of cirrhosis and associated complications, these patients often have recurrent hospital admissions. Along with increasing admissions for cirrhosis, the economic burden for management of cirrhosis also appears to be on the rise[5]. Readmission rates have been reported to range from 20% to 50% in patients with cirrhosis[7–9]. Reliance on readmission rate would not depict a clear picture of healthcare delivery in this chronic condition without simultaneously studying the impact on patient mortality[10]. Currently, studies assessing trends in multiple outcomes such as 30-day inpatient mortality, readmission rates, and hospitalization cost in patients with cirrhosis (i.e., simultaneously) are lacking. Furthermore, CC patients remain understudied.
Here, we hypothesized that as a result of improved inpatient care of CC and DC patients, there would be a corresponding trend of reduced 30-day inpatient mortality, 30-day readmission, associated cost of hospitalization, and 30-day readmission. We also aimed to examined factors associated with 30-day mortality and all-cause readmission.
Material and Methods
Data Source
The Nationwide Readmission Database (NRD) from the year 2010 to 2016 was used for this study. NRD is a part of a family of databases and software tools developed by the Agency for Healthcare Research and Quality’s Healthcare Cost and Utilization Project (AHRQ-HCUP)[11]. The NRD is a unique and powerful database designed to support various types of analyses of national readmission rates. It is the largest publicly available all-payer inpatient readmission database with approximately 17 million discharges each year, and when weighted, it estimates roughly 36 million discharges. NRD contains data from 22 to 27 geographically dispersed States (varied as per year of data), accounting for more than 50 percent of the total U.S. resident population and all U.S. hospitalizations[11]. The NRD contains more than 100 clinical and nonclinical variables for each hospital stay, including variables essential to readmission analyses. It includes up to 25 discharge diagnoses and 15 procedures for each patient using the International Classification of Diseases, Ninth and Tenth Revision, Clinical Modification (ICD-9/10-CM; varied based on year of data) codes. The patient record has unique patient linkage numbers through which patients’ admissions to any hospital within a particular state can be tracked.
NRD is a publicly available database that contains completely de-identified data without compromising patient confidentiality, and so this study was exempt from Institutional Review Board review. User data agreement with the Agency for Healthcare Research and Quality for the use of NRD database was completed[12].
Study Population
Patients with a diagnosis of cirrhosis were identified using ICD-9-CM codes (for NRD year 2010-2015) and ICD-10-CM codes (for NRD year 2015-2016) (Appendix Table 1). If a patient with cirrhosis had at least one feature of decompensation (ascites, HE, SBP, or variceal bleeding) as per Baveno IV criteria[13], that patient was categorized as having DC. The rest of the patients were classified as CC. Patients younger than 18 years of age, missing information on age, gender, and mortality were excluded. As one of the primary objectives of this study was to calculate the 30-day readmission rate, patients with missing information on the length of stay (LOS), index hospitalization in December, death during the index hospitalization, and elective readmission were excluded during readmission rate analysis (Figure 1). The reason that we excluded discharges in December is that it is not possible to track the patient during the following year because the patient tracking number changes at the beginning of each new year.
Figure 1. Study cohort.
The consort diagram depicts the numbers of patients in the study population, including selection and exclusion criteria from readmission analysis for compensated & decompensated cirrhosis patients.
Study Outcomes
For both for CC and DC patients, the primary outcomes of the study were 30-day all-cause mortality, 30-day all-cause readmission, and hospitalization cost associated with index hospitalization and 30-day readmissions. Secondary outcomes included: 1) factors associated with 30-day all-cause mortality, 2) factors associated with 30-day all-cause readmission.
Definitions of Variables
NRD pre-defined variables were used to identify each patient’s age (in years), gender (male or female), comorbidities, LOS, inpatient mortality, hospitalization costs, primary insurance type, disposition, and admission day[14]. The Charlson comorbidity index (CCI) was used to assess comorbidity burden[15]. CCI score was modified by excluding the score for mild and severe liver disease. Primary discharge diagnoses, complications of cirrhosis, procedures on index hospitalization, and etiologies for 30-readmissions were identified using ICD-9/10-CM diagnoses/procedure codes. The early discharge was defined as a length of stay of less than three days. Hospital charge represents the amount that hospital billed for during that hospital stay, but do not reflect the actual cost of care. With the hospital-specific cost to charge ratio files[16] provided by HCUP-AHRQ mean cost of hospitalization, as well as the total cost of hospitalization over the years, was derived. The cost of hospitalization was inflation-adjusted with an inflation calculator available on the Bureau of labor statistics[17]. Using the same procedure, hospital charges associated with the 30-day readmission were derived.
Statistical analysis
Descriptive statistics were utilized to describe the study population with categorical variables reported as percentages and numbers, while continuous variables were reported as mean. Bivariate group comparisons between readmitted patients and those who were not readmitted within 30 days of hospital discharges were made using weighted χ2 analysis for categorical variables and weighted Student t-test for continuous variables. Unplanned readmissions were calculated using the variables “NRD_visitlink” and “NRD_DaysToEvent” which have been created within NRD to track the patients. Reasons for readmissions were identified, tallying primary discharge diagnosis for each readmitted patient. Multivariable logistic regression analysis was used to obtain adjusted odds ratios to analyze the predictors of 30-day all-cause mortality and 30-day all-cause readmission. Covariates with P < 0.1 on univariate analysis were entered into the model. The results of multivariable logistic regression were reported as adjusted odds ratio (aOR) with 95% CI (confidence intervals). NRD allows calculating national estimates based on its complex sampling designs, which include stratification, clustering, and weighting. These were considered to obtain estimates for the entire population in the United States of hospitalized cirrhosis patients. P-values were considered significant a priori as <0.05. All analyses were performed with SAS statistical software (version 9.4; SAS Institute Inc, Cary, NC).
Results
Within the dataset, a total of 3,396,998 patients with cirrhosis were hospitalized between 2010 to 2016 after excluding patients (Figure 1). Out of all cirrhosis patients, 1,683,004 (49.8%) had a documented decompensation event on presentation (Figure 1). After excluding patients less than 18 year old, missing data on LOS, death during initial hospitalization and hospitalization in month of December for each year, we had total of 3,374,038 patients with cirrhosis and out of those 1,338,233 were decompensated.
As expected, the mean age at index admission was nearly 60 years, and there were more males than females, consistent with cirrhosis population (Appendix Table 2). Patients with DC were even more likely to be male. CC and DC cirrhosis patients had comparable CCI’s and proportions of weekend admissions. Also, as expected, 30-day inpatient mortality was over two-fold higher in DC (11.3%) compared to CC (5.4%) patients. Interestingly, the proportion of patients having an early discharge (within three days) was higher for DC than CC (66% vs. 46%). DC patients were more likely to have alcoholic cirrhosis than CC patients (51% vs. 39%). The prevalence of HCV and HBV was similar among the two groups. In DC group, the most common decompensating event was ascites (83.2%) followed by HE (30%). Some 50% of DC patients had paracentesis, and over 20% had an EGD. The median LOS was one day longer (5 vs. 4 days), and the cost of hospitalization was nearly $ 1,000 higher ($ 10,635 vs. $ 9021) in DC patients compared to CC patients.
We also examined differences in patient demographics, clinical features, and hospital resource utilization for CC and DC patients as a function of 30-day readmission status (Table 1). In both groups, the proportion of the middle-age population was higher, and the proportion of elderly patients was lower in readmitted cohorts. Patients insured by Medicaid had a higher proportion of readmission. In contrast, patients with private insurance had less frequent readmission than patients without readmission (in both CC and DC cohorts).
Table 1.
Baseline characteristics
| Compensated cirrhosis | Decompensated cirrhosis | |||
|---|---|---|---|---|
| Variables | 30-day readmission (n=354,707) |
No 30-day readmission (n=1,631,098) |
30-day readmission (n=388,864) |
No 30-day readmission (n=999,369) |
| Age | ||||
| Mean (years) | 59 | 60 | 58 | 59 |
| 18-39 | 17,659 (5) | 76,319 (5) | 20,707 (5) | 46,991 (5) |
| 40-64 | 231,498 (65) | 1,003,353 (62) | 267,752 (69) | 650,065 (65) |
| ≥65 | 105,549 (30) | 551,255 (34) | 100,405 (26) | 302,313 (30) |
| Gender | ||||
| Male | 209,243 (59) | 956,979 (59) | 244,156 (63) | 627,327 (63) |
| Female | 145,464 (41) | 674,119 (41) | 144,708 (37) | 372,042 (37) |
| CCI | ||||
| Mean | 1.8 | 1.7 | 1.6 | 1.6 |
| 0 | 100,344 (28) | 489,458 (30) | 132,594 (34) | 351,778 (35) |
| 1 | 87,586 (25) | 429,115 (26) | 85,282 (22) | 228,934 (23) |
| 2 | 63,932 (18) | 288,740 (18) | 68,927 (18) | 167,362 (17) |
| ≥3 | 102,843 (29) | 423,784 (26) | 102,062 (26) | 251,296 (25) |
| Weekend admission | 82,875 (23) | 357,856 (82) | 89,101 (23) | 221,557 (22) |
| Insurance type | ||||
| Medicare | 174,918 (49) | 808,674 (50) | 165,018 (42) | 434,956 (44) |
| Medicaid | 98,804 (28) | 357,250 (22) | 114,686 (30) | 234,472 (24) |
| Private | 51,784 (15) | 296,924 (18) | 74,362 (19) | 216,780 (22) |
| Self-pay | 14,748 (4) | 86,735 (5) | 18,150 (5) | 61,179 (6) |
| Other | 14,451 (4) | 81,515 (5) | 16,648 (4) | 51,982 (5) |
| Early discharge | 157,669 (44) | 752,567 (46) | 258,467 (67) | 658,105 (66) |
| Disposition | ||||
| Routine | 212,917 (72) | 1,043,925 (77) | 238,840 (73) | 602,382 (73) |
| TransferØ | 4555 (2) | 21,155 (2) | 5595 (2) | 19,404 (2) |
| Home with HH | 62,020 (21) | 256,720 (19) | 70,636 (22) | 181,109 (22) |
| AMA | 14,731 (5) | 39,601 (3) | 12,963 (4) | 18,207 (2) |
| Etiology of cirrhosis | ||||
| Alcoholic | 149,699 (42) | 619,350 (38) | 206,215 (53) | 507,394 (51) |
| Non-alcoholic | 205,008 (58) | 1,011,748 (62) | 182,649 (47) | 491,975 (49) |
| Concurrent hepatitis infection | ||||
| HBV | 7362 (2) | 28,588 (2) | 8687 (2) | 18,972 (2) |
| HCV | 122,137 (34) | 481,428 (30) | 125,024 (32) | 276,301 (28) |
| Features of decompensation | ||||
| Ascites | 336,312 (87) | 818,793 (82) | ||
| SBP | 26,742 (7) | 63,738 (6) | ||
| HE | 127,222 (33) | 284,723 (29) | ||
| Variceal bleeding | 37,043 (10) | 139,844 (14) | ||
| HRS | 24,857 (6) | 59,467 (6) | ||
| HCC | 21,461 (6) | 57,358 (6) | ||
| Resource utilization | ||||
| Interventions at index hospitalization | ||||
| Paracentesis | 200,934 (52) | 428,406 (43) | ||
| TIPS | 7586 (2) | 17,864 (2) | ||
| EGD | 125,331 (17) | 453,643 (17) | 77,895 (20) | 239,064 (24) |
| Hemodialysis | 45,441 (6) | 112,094 (4) | 25,045 (6) | 51,639 (5) |
| Median Length of stay (cleaned, days) | 4 | 4 | 5 | 5 |
| Median Cost of hospitalization ($) | 8649 | 9105 | 10,304 | 10,768 |
SBP: Spontaneous bacterial peritonitis; HE: Hepatic encephalopathy; HRS: Hepatorenal syndrome; HCC: Hepatocellular carcinoma; EGD: Esophagogastroduodenoscopy; TIPS: Transjugular intrahepatic portosystemic shunt; AMA: Against medical advice
transfer to skilled nursing facility or acute care hospital
Numbers presented in ( ) represents percentage
Trends of 30-day inpatient mortality, readmission rate, cost of hospitalization, cost of readmission, and index length of stay
Interestingly, the total number of discharges for DC patients increased over the study period from 416 to 730 per 100,000 discharges (Figure 2A). The overall 30-day inpatient mortality rate for CC and DC patients decreased from 6.1% to 4%, and from 12.1% to 10.6%, respectively (Figure 2B). Readmission rates for DC patients over the seven-year study period, even after adjusting for age, sex, and comorbidities, remained constant at approximately 28% for DC patients (Figure 3). However, for CC patients, readmission rates dropped from 18% to 16%. We found that in patients readmitted, 50% and 75% of DC patients were readmitted by day 11 and day 19, respectively (Appendix Figure 1). Findings were similar in CC patients (Appendix Figure 2). The total cost for hospitalization for patients with DC increased substantially, by nearly $3 billion over the study period (Figure 4A), while the total cost of hospitalization for CC patients was only slightly increased over the same time period. The cost for patients being readmitted within 30 days also increased substantially for DC patients but remained nearly stable for CC patients (Figure 4B). On a patient level, the median cost for index hospitalization remained nearly unchanged, while the cost of readmission increased by a greater degree and was higher for DC than CC patients (Figure 4C & D). The median length of stay for index hospitalization remained nearly stable in both DC and CC groups (Appendix Figure 3).
Figure 2. Trends in discharges & mortality.
In A) is shown the number of patients discharged with compensated and decompensated cirrhosis from the year 2010 to 2016. In B) is shown the trend of 30-day all-cause inpatient mortality for compensated cirrhosis and decompensated cirrhosis patients from the year 2010 to 2016.
Figure 3. Trends in readmission.
The figure depicts 30-day all-cause readmission rate for compensated and decompensated cirrhosis from the year 2010 to 2016.
Figure 4. Hospital costs.
In A), total costs of the index hospitalization for compensated cirrhosis and decompensated cirrhosis patients from 2010 to 2016 are shown (in US dollars, adjusted for inflation). In B), the total costs of all-cause readmission for compensated cirrhosis and decompensated cirrhosis patients from 2010 to 2016 are shown (in US dollars, adjusted for inflation). In C), the median cost of hospitalization for compensated cirrhosis and decompensated cirrhosis patients from 2010 to 2016 is shown (in US dollars, adjusted for inflation). In D), the median cost of all-cause readmission for compensated cirrhosis and decompensated cirrhosis patients from 2010 to 2016 is shown (in US dollars, adjusted for inflation).
Causes of readmission
We also investigated the causes of 30-day readmission by identifying the primary discharge diagnoses of readmitted patients in CC and DC groups (Figure 5). Etiologies related to the gastroenterological system were identified in the majority of the readmitted population (55% in DC group and 41.5% in CC group), followed by infectious disease (20% in DC group and 18% in CC group). Overall, hepatic encephalopathy, infections, and ascites were the top 3 discharge diagnoses in the DC cohort. Interestingly, out of all CC patients, 7.2% (142, 809) were readmitted with decompensation (54% ascites, 54% HE, 4% SBP, and 7% variceal bleeding; these groups were not mutually exclusive), and out of all readmitted CC patients, almost 1/3rd of them were decompensated on first 30-day re-admission.
Figure 5. Causes of readmission.
The proportion of compensated and decompensated cirrhosis patients with a specific primary discharge diagnosis were coded for the readmission and sorted by the organ system involved. For example, the gastrointestinal system represents all readmissions with diagnoses pertaining to gastrointestinal disease, including, but not limited to, ascites, hepatic encephalopathy, GI bleeding, and spontaneous bacterial peritonitis.
Factors associated with 30-day inpatient mortality and readmission
Compensated cirrhosis
Alcoholic cirrhosis and HBV infection were the etiologies of cirrhosis associated with the greatest increase in inpatient mortality (Table 2). As expected, CCI was strongly associated with inpatient mortality. Medicare insurees had 15% lower inpatient mortality and 30% higher 30-day readmission than commercial insurees. Weekend admission did not impact inpatient mortality. Discharge within three days was associated with 25% lower inpatient mortality without increasing the odds of readmission. Transfer to a skilled nursing facility or acute care hospital was associated with nearly 2.5 -fold higher inpatient mortality but did not increase the odds of readmission. AMA discharges were associated with 67% higher odds of being readmitted within 30 days.
Table 2.
Predictors of 30-day mortality and 30-day readmission in compensated cirrhosis patients
| 30- day all-cause inpatient mortality | 30-day all-cause readmission | |||
|---|---|---|---|---|
| Variables | Odds Ratio (95% CI) | P – value | Odds Ratio (95% CI) | P – value |
| Age | ||||
| 18-39 | 0.68 (0.58 – 0.81) | <.0001 | 1.38 (1.32 – 1.44) | <.0001 |
| 40-64 | 0.85 (0.79 – 0.91) | <.0001 | 1.25 (1.22 – 1.28) | <.0001 |
| ≥65 | Reference | |||
| Gender | ||||
| Male | 0.97 (0.91 – 1.02) | 0.315 | 0.95 (0.93 – 0.97) | <.0001 |
| Female | ||||
| Etiology of cirrhosis | ||||
| Alcoholic | 1.27 (1.20 – 1.34) | <.0001 | 1.27 (1.25 – 1.29) | <.0001 |
| Non-alcoholic | Reference | |||
| Concurrent hepatitis infection | ||||
| Hepatitis B | 1.38 (1.14 – 1.67) | 0.001 | 1.16 (1.10 – 1.23) | <.0001 |
| Hepatitis C | 1.07 (1.01 – 1.12) | 0.022 | 1.21 (1.19 – 1.23) | <.0001 |
| None | Reference | |||
| Charlson co-morbidity index | ||||
| 0 | 0.99 (0.91 – 1.08) | 0.942 | 1.01 (0.99 – 1.03) | 0.157 |
| 1 | 1.37 (1.26 – 1.50) | <.0001 | 1.12 (1.20 – 1.15) | <.0001 |
| 2 | 1.80 (1.67 – 1.94) | <.0001 | 1.28 (1.25 – 1.31) | <.0001 |
| ≥3 | Reference | |||
| Day of admission | ||||
| Weekend | 1.00 (0.94 – 1.06) | 0.915 | 1.08 (1.06 – 1.10) | <.0001 |
| Weekday | Reference | |||
| Primary payer | ||||
| Medicare | 0.85 (0.77 – 0.93) | <.0001 | 1.33 (1.29 – 1.36) | <.0001 |
| Medicaid | 1.06 (0.97 – 1.16) | 0.176 | 1.48 (1.44 – 1.52) | <.0001 |
| Self–pay | 0.85 (0.72 – 1.00) | 0.054 | 0.93 (0.90 – 0.97) | 0.002 |
| Other | 0.66 (0.56 – 0.77) | <.0001 | 1.00 (0.96 – 1.04) | 0.811 |
| Private insurance | Reference | |||
| Discharge time | ||||
| Early (<4 days) | 0.72 (0.68 – 0.76) | <.0001 | 0.96 (0.94 – 0.97) | <.0001 |
| Late (≥ 4 days) | Reference | |||
| Disposition | ||||
| Against medical advice | 1.34 (1.17 – 1.53) | <.0001 | 1.67 (1.61 – 1.73) | <.0001 |
| TransferØ | 2.45 (2.08 – 2.87) | <.0001 | 1.06 (0.99 – 1.13) | 0.053 |
| Home health care | 1.56 (1.46 – 1.67) | <.0001 | 1.18 (1.16 – 1.21) | <.0001 |
| Routine | Reference | |||
transfer to skilled nursing facility or acute care hospital
Decompensated cirrhosis
Out of all decompensating features, ascites and HE were the most strongly associated with inpatient mortality and readmission (ascites was associated with 70% higher odds of inpatient of mortality and 30% higher odds of 30-day readmission, while HE was associated with nearly 30% higher inpatient mortality and almost 50% higher odds of 30-day readmission) (Table 3). Notably, SBP and variceal bleeding did not appear to be associated with increased inpatient mortality. Odds of inpatient mortality and 30-day readmission were almost 40% higher in patients receiving paracentesis during admission. TIPS, although associated with an increased odds of readmission (by 13%), was associated with 30% lower inpatient mortality. Similar to CC patients, weekend admissions did not have mortality differences compared to weekday admission but had slightly increased odds of 30-day readmission. AMA discharges had 50% higher odds of inpatient mortality and 70% higher odds of 30-day readmission than patients who were discharged home.
Table 3.
Predictors of 30-day mortality and 30-day readmission in decompensated cirrhosis patients
| 30- day all-cause inpatient mortality | 30-day all-cause readmission | |||
|---|---|---|---|---|
| Variables | Odds Ratio (95% CI) | P – value | Odds Ratio (95% CI) | P – value |
| Age | ||||
| 18-39 | 0.80 (0.71 – 0.88) | <.0001 | 1.39 (1.33 – 1.45) | <.0001 |
| 40-64 | 0.95 (0.89 – 1.01) | 0.115 | 1.22 (1.20 – 1.25) | <.0001 |
| ≥65 | Reference | |||
| Gender | ||||
| Male | 1.06 (1.02 – 1.10) | 0.008 | 0.97 (0.95 – 0.99) | 0.002 |
| Female | Reference | |||
| Etiology of cirrhosis | 1.14 (1.10 – 1.20) | < .0001 | 1.10 (1.08 – 1.12) | <.0001 |
| Alcoholic | 1.14 (1.10 – 1.20) | < .0001 | 1.10 (1.08 – 1.12) | <.0001 |
| Non–alcoholic | Reference | |||
| Concurrent hepatitis infection | ||||
| Hepatitis B | 1.44 (1.30 – 1.60) | <.0001 | 1.23 (1.17 – 1.20) | <.0001 |
| Hepatitis C | 1.08 (1.03 – 1.12) | 0.001 | 1.27 (1.25 – 1.30) | <.0001 |
| None | Reference | |||
| Charlson co-morbidity index | ||||
| 0 | 0.95 (0.89 – 1.00) | 0.066 | 1.03 (1.01 – 1.05) | 0.003 |
| 1 | 1.10 (1.04 – 1.18) | 0.001 | 1.14 (1.11 – 1.17) | <.0001 |
| 2 | 1.13 (1.06 – 1.20) | < .0001 | 1.18 (1.15 – 1.20) | <.0001 |
| ≥3 | Reference | |||
| Decompensating feature a | ||||
| Ascites | 1.70 (1.58 – 1.84) | <.0001 | 1.30 (1.26 – 1.32) | <.0001 |
| Hepatic encephalopathy | 1.27 (1.21 – 1.33) | < .0001 | 1.47 (1.44 – 1.50) | <.0001 |
| Spontaneous bacterial peritonitis | 1.07 (0.99 – 1.16) | 0.058 | 0.97 (0.94 – 1.00) | 0.104 |
| Variceal bleeding | 1.02 (0.94 – 1.10) | 0.542 | 0.83 (0.81 – 0.85) | <.0001 |
| Hepatocellular carcinoma | ||||
| Yes | 1.65 (1.51 – 1.80) | <.0001 | 0.90 (0.86 – 0.93) | <.0001 |
| No | Reference | |||
| Hepatorenal syndrome | ||||
| Yes | 1.51 (1.41 – 1.61) | <.0001 | 1.02 (0.97– 1.05) | 0.390 |
| No | Reference | |||
| Day of admission | ||||
| Weekend | 1.02 (0.97 – 1.07) | 0.462 | 1.06 (1.04 – 1.08) | <.0001 |
| Weekday | Reference | |||
| Primary payer | ||||
| Medicare | 1.01 (0.94 – 1.07) | 0.848 | 1.21 (1.18 – 1.24) | <.0001 |
| Medicaid | 1.14 (1.07 – 1.21) | <.0001 | 1.37 (1.33 – 1.40) | <.0001 |
| Self-pay | 1.05 (0.95 – 1.15) | 0.344 | 0.86 (0.83 – 0.90) | <.0001 |
| Other | 0.95 (0.86 – 1.05) | 0.350 | 0.96 (0.92 – 1.00) | 0.083 |
| Private insurance | Reference | |||
| Discharge time | ||||
| Early (<4 days) | 0.81 (0.77 – 0.84) | <.0001 | 1.07 (1.05 – 1.09) | <.0001 |
| Late (≥ 4 days) | Reference | |||
| Disposition | ||||
| Against medical advice | 1.51 (1.36 – 1.68) | < .0001 | 1.70 (1.63 – 1.76) | <.0001 |
| Transferb | 1.31 (1.13 – 1.52) | < .0001 | 0.71 (0.65 – 0.76) | <.0001 |
| Home health care | 1.21 (1.15 – 1.28) | <.0001 | 0.93 (0.91 – 0.95) | <.0001 |
| Routine | Reference | |||
| Procedures on index hospitalization c | ||||
| Paracentesis | 1.38 (1.33 – 1.45) | < .0001 | 1.37 (1.34 – 1.39) | <.0001 |
| TIPS | 0.71 (0.57 – 0.86) | <.0001 | 1.13 (1.06 – 1.20) | <.0001 |
| EGD | 0.95 (0.90 – 1.00) | 0.099 | 0.91 (0.90 – 0.93) | <.0001 |
| Hemodialysis | 1.08 (1.01 – 1.17) | 0.035 | 1.14 (1.10 – 1.18) | <.0001 |
compared to absence of respective decompensating feature
transfer to skilled nursing facility or acute care hospital
compared to absence of respective procedure
TIPS: Transjugular Intrahepatic Portosystemic Shunt; EGD: Esophagogastroduodenoscopy
Discussion
Our study assessed three crucial aspects of healthcare delivery in patients with cirrhosis, including the following: mortality, readmission, and costs. We have shown that the in-hospital mortality rate appears to be declining for both CC and DC patients, and there has been a decline in the 30-day readmission in CC. Meanwhile, the financial burden to the healthcare system for both CC and DC patients is on the rise - despite no change in the hospital length of stay. Finally, most readmissions for CC and DC are attributable to gastrointestinal and infectious diseases.
Previous studies suggesting improvements in mortality in cirrhosis patients have been limited to patients with complications associated with cirrhosis[18–22]. Our study is the first to our knowledge that has reported on the trend of 30-day inpatient mortality rate for both CC and DC. As might be expected, mortality was higher in DC patients (11.3% vs. 5.4%) compared to CC patients. We specifically noted a decrease in inpatient mortality for both CC and DC patients, perhaps due to improved efficiency in inpatient care delivery, but maybe also due to earlier discharge. The latter possibility is supported by a study of 126 VA hospitals that reported 30% lower inpatient mortality, but also 10% increase in 30-day post-discharge mortality among cirrhosis patients during the study period – consistent with a shift of the mortality burden to the outpatient setting[23].
Readmission following initial hospitalization has been an important quality and accountability metric for healthcare delivery. Importantly, For a chronic condition such as cirrhosis, it may serve as a surrogate for post-discharge care. A recent retrospective study using state inpatient data from the year 2009 to 2013 showed that the 30-day readmission rate did not change over that time period, but that the rate of readmission was higher in cirrhosis patients (around 30%) than those without cirrhosis[24]. Decompensation in cirrhosis patients predicts a worse outcome and also suggests the progression of underlying disease process[6]. These patients are often challenging to manage, and they are frequently hospitalized as a result of complications of cirrhosis. Historically, the 30-day readmission rate in DC patients has been reported anywhere from 20 to 50%[7, 8, 25, 26]. Data from our study are consistent with this finding, showing that the readmission rate was 28% (and has remained unchanged). But it remains higher compared to CC patients. Interestingly, we noted a decrease in readmission for CC patients, which may be a result of improvements in the quality of care in these patients.
An important finding of this study is that the cost of care for patients with cirrhosis appears to be increasing. Our study is unique in this aspect because not only did we examine the total cost of initial hospitalization, but also we evaluated the financial burden imposed by unplanned 30-day readmission. Consistent with our data is a study that assessed the economic burden of cirrhosis using the National Inpatient Sample (NIS), which reported a 33% and 25% increase in the total cost of hospitalization for DC and CC patients, respectively, with total cost amounting to some $4.5 billion for DC and $2.8 billion for CC. Our analysis revealed rising hospitalization costs for DC and decreasing hospitalization costs for CC reaching $5.3- and $4- billion, respectively, in 2016. Moreover, we also report an almost two-fold rise in total cost attributed to readmission for DC. The mean cost of hospitalization and readmission is higher for DC than CC, suggesting these patients have greater resource utilization once admitted, almost certainly in the form of treatments used for cirrhotics (such as antibiotics, octreotide, albumin) and procedures (endoscopy, paracentesis, dialysis). Of note, the economic burden in this patient population seems to be driven by complications of portal hypertension and in hospital procedures[5]. Finally, readmission risk models have been proposed[8, 25], leading to interventions that curb readmission costs.
Notably, over half of readmissions occurred shortly after discharge (12 and 11 days for DC and CC patients, respectively). This finding raises the possibility that early follow-up visits after discharge could prevent unnecessary re-hospitalization. We also studied the cause of readmission – finding that gastrointestinal and infectious diseases were the top two systems leading to readmission in both CC and DC. Thus, it is likely that focus on complications of portal hypertension, such as variceal bleeding, worsening ascites (in some instances with SBP), hepatic encephalopathy may prove beneficial. Further, preemptive attention to these types of disorders, perhaps with a discharge check list[27] and chronic disease management care team[28] would be helpful.
We have also examined clinical variables associated with inpatient mortality and 30-day readmission. An obvious demographic risk factor would appear to be ongoing ethanol use. Consistent with this possibility is that patients with alcoholic cirrhosis had higher readmission rate and 30-day mortality compared to non-alcoholic cirrhosis in both CC and DC groups. Although we cannot definitively determine whether ongoing alcohol use in some patients was important, this remains a possibility and could be important in pushing CC patients into the DC state. Concurrent hepatitis B infection was associated with higher 30-day inpatient mortality and readmission risk, emphasizing the need for treating such patients. Ascites and HE are common causes of readmission and were associated with higher 30-day readmission and mortality in DC cohort, consistent with their being important causes of decompensation.
Certain inpatient procedures are vital in caring for cirrhosis patients. Paracentesis is often performed in patients with a new diagnosis of ascites or those with pre-existing cirrhosis as a workup of SBp[29]; these clinical points may explain the higher readmission and mortality rates associated with this procedure. TIPS is often performed in patients with recurrent variceal bleeding or for those with refractory ascites[17]. Interestingly, TIPS was associated with a lower 30-day inpatient mortality rate, suggesting that this procedure may have benefit in certain patients. However, inpatient TIPS was associated with an increased risk of readmissions, perhaps due to an increased risk of hepatic encephalopathy following this procedure. EGD is another critical procedure mainly for decompensated cirrhosis patients who present with upper GI bleeding. We noticed reduced readmission risk after EGD, raising the possibility that EGD is associated with the identification and treatment of bleeding varices.
We recognize the limitations of this study, several of which are inherent in using administrative data such as done here. First, our analysis relied on ICD-9/10 CM codes for diagnosis and procedures – and thus, we cannot exclude the possibility of miscoding. However, we have previously demonstrated that specific ICD-9 codes have a high sensitivity and specificity for the detection of cirrhosis[30]. Additionally, it should be noted that CMS uses the same coding methodology to identify readmissions. NRD lacks data on medications or laboratory values, so we could not account for the effects of certain medications and/or medical compliance of patients in our analysis. Also, without MELD or Child-Turcot-Pugh scores, we could not control the severity of the underlying liver disease. The NRD only has data for one calendar year and patients can not be tracked over next year if admitted in latter part of year. Thus, we excluded patients admitted in December in our study. Implications of exclusion of patients on readmission rate has not been significant when studying 30-day readmission rate. However, these weaknesses are likely offset by the large number of patients included in the data and the diverse population from which the data are drawn. NRD has been used to study readmissions in several medical and surgical conditions previously[25, 31] and it is the largest available US national database available.
In conclusion, we have provided updated mortality data about patients admitted with cirrhosis, and new information about readmission costs, causes of readmission, and factors leading to readmissions in CC and DC patients. Our data suggest that inpatient care of CC and DC patients has likely improved in the last decade, as evident from declining 30-day inpatient mortality. However, the financial burden of care is increasing, with readmissions being costly. Finally, our data point to the possibility that preventive measures at discharge and timely post-discharge follow up could lead to reduced hospital costs.
Supplementary Material
Figure A1. Timing of readmission for compensated cirrhosis patients. The histogram depicts the percentage of decompensated cirrhosis patients readmitted over a 30 day period with bold arrows marking 50% all 30-day readmission on day 11.
Figure A2. Timing of readmission for decompensated cirrhosis patients. The histogram depicts the percentage of decompensated cirrhosis patients readmitted patients over a 30 day period with bold arrows marking 50% all 30-day readmission on day 12
Figure A3. Length of stay during the index hospitalization. The graph shows the trends in median length of stay for compensated and decompensated cirrhosis patients
Appendix Table 1. ICD-9 & ICD-10 codes for diagnoses and procedures
Appendix Table 2. Baseline characteristics of index patients
Financial Funding:
DCR was supported in part by the National Institutes of Health, P30 DK123704.
Footnotes
Conflict of Interest: The Authors have no conflict of interest to declare.
Ethics approval statement: This study was exempt from IRB approval due to use of de-identified patient database.
References
- 1.Asrani SK, Larson JJ, Yawn B, et al. Underestimation of liver-related mortality in the United States. Gastroenterology. 2013;145:375–382 e371–372. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Kim WR, Brown RS Jr., Terrault NA, et al. Burden of liver disease in the United States: summary of a workshop. Hepatology. 2002;36:227–242. [DOI] [PubMed] [Google Scholar]
- 3.Neff GW, Duncan CW and Schiff ER. The current economic burden of cirrhosis. Gastroenterol Hepatol (N Y). 2011;7:661–671. [PMC free article] [PubMed] [Google Scholar]
- 4.Allen AM, Kim WR, Moriarty JP, et al. Time trends in the health care burden and mortality of acute on chronic liver failure in the United States. Hepatology. 2016;64:2165–2172. [DOI] [PubMed] [Google Scholar]
- 5.Desai AP, Mohan P, Nokes B, et al. Increasing Economic Burden in Hospitalized Patients With Cirrhosis: Analysis of a National Database. Clin Transl Gastroenterol. 2019;10:e00062. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.D’Amico G, Garcia-Tsao G and Pagliaro L. Natural history and prognostic indicators of survival in cirrhosis: a systematic review of 118 studies. J Hepatol. 2006;44:217–231. [DOI] [PubMed] [Google Scholar]
- 7.Scaglione SJ, Metcalfe L, Kliethermes S, et al. Early Hospital Readmissions and Mortality in Patients With Decompensated Cirrhosis Enrolled in a Large National Health Insurance Administrative Database. J Clin Gastroenterol. 2017;51:839–844. [DOI] [PubMed] [Google Scholar]
- 8.Volk ML, Tocco RS, Bazick J, et al. Hospital readmissions among patients with decompensated cirrhosis. Am J Gastroenterol. 2012;107:247–252. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Tapper EB, Halbert B and Mellinger J. Rates of and Reasons for Hospital Readmissions in Patients With Cirrhosis: A Multistate Population-based Cohort Study. Clin Gastroenterol Hepatol. 2016;14:1181–1188 e1182. [DOI] [PubMed] [Google Scholar]
- 10.Henry Z Readmissions for Cirrhosis Within the Healthcare Readmissions Reduction Program: A Hidden Challenge. Am J Gastroenterol. 2019;114:1419–1420. [DOI] [PubMed] [Google Scholar]
- 11.Dobrzycka M, Spychalski P, Łachiński AJ, et al. Statins and Colorectal Cancer - A Systematic Review. Exp Clin Endocrinol Diabetes. 2020;128:255–262. [DOI] [PubMed] [Google Scholar]
- 12.Nationwide Data Use Agreement. Healthcare Cost and Utilization Project (HCUP). September 2019. Agency for Healthcare Research and Quality R, MD. www.hcup-us.ahrq.gov/team/NationwideDUA.jsp. Accessed on 02/26/2020. [Google Scholar]
- 13.de Franchis R Evolving consensus in portal hypertension. Report of the Baveno IV consensus workshop on methodology of diagnosis and therapy in portal hypertension. J Hepatol. 2005;43:167–176. [DOI] [PubMed] [Google Scholar]
- 14.Shah SC, Glass J, Giustino G, et al. Statin Exposure Is Not Associated with Reduced Prevalence of Colorectal Neoplasia in Patients with Inflammatory Bowel Disease. Gut Liver. 2019;13:54–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Simard M, Sirois C and Candas B. Validation of the Combined Comorbidity Index of Charlson and Elixhauser to Predict 30-Day Mortality Across ICD-9 and ICD-10. Med Care. 2018;56:441–447. [DOI] [PubMed] [Google Scholar]
- 16.Tanaka S, Nemoto Y, Takei Y, et al. High-fat diet-derived free fatty acids impair the intestinal immune system and increase sensitivity to intestinal epithelial damage. Biochem Biophys Res Commun. 2020;522:971–977. [DOI] [PubMed] [Google Scholar]
- 17.Copelan A, Kapoor B and Sands M. Transjugular intrahepatic portosystemic shunt: indications, contraindications, and patient work-up. Semin Intervent Radiol. 2014;31:235–242. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Stepanova M, Mishra A, Venkatesan C, et al. In-hospital mortality and economic burden associated with hepatic encephalopathy in the United States from 2005 to 2009. Clin Gastroenterol Hepatol. 2012;10:1034–1041 e1031. [DOI] [PubMed] [Google Scholar]
- 19.Singal AK, Salameh H and Kamath PS. Prevalence and in-hospital mortality trends of infections among patients with cirrhosis: a nationwide study of hospitalised patients in the United States. Aliment Pharmacol Ther. 2014;40:105–112. [DOI] [PubMed] [Google Scholar]
- 20.Mishra A, Otgonsuren M, Venkatesan C, et al. The inpatient economic and mortality impact of hepatocellular carcinoma from 2005 to 2009: analysis of the US nationwide inpatient sample. Liver Int. 2013;33:1281–1286. [DOI] [PubMed] [Google Scholar]
- 21.Vergara M, Cleries M, Vela E, et al. Hospital mortality over time in patients with specific complications of cirrhosis. Liver Int. 2013;33:828–833. [DOI] [PubMed] [Google Scholar]
- 22.Schmidt ML, Barritt AS, Orman ES, et al. Decreasing mortality among patients hospitalized with cirrhosis in the United States from 2002 through 2010. Gastroenterology. 2015;148:967–977 e962. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Kanwal F, Tansel A, Kramer JR, et al. Trends in 30-Day and 1-Year Mortality Among Patients Hospitalized With Cirrhosis From 2004 to 2013. Am J Gastroenterol. 2017;112:1287–1297. [DOI] [PubMed] [Google Scholar]
- 24.Rosenblatt R, Cohen-Mekelburg S, Shen N, et al. Cirrhosis as a Comorbidity in Conditions Subject to the Hospital Readmissions Reduction Program. Am J Gastroenterol. 2019;114:1488–1495. [DOI] [PubMed] [Google Scholar]
- 25.Mumtaz K, Issak A, Porter K, et al. Validation of Risk Score in Predicting Early Readmissions in Decompensated Cirrhotic Patients: A Model Based on the Administrative Database. Hepatology. 2019;70:630–639. [DOI] [PubMed] [Google Scholar]
- 26.Berman K, Tandra S, Forssell K, et al. Incidence and predictors of 30-day readmission among patients hospitalized for advanced liver disease. Clin Gastroenterol Hepatol. 2011;9:254–259. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Tapper EB, Finkelstein D, Mittleman MA, et al. A Quality Improvement Initiative Reduces 30-Day Rate of Readmission for Patients With Cirrhosis. Clin Gastroenterol Hepatol. 2016;14:753–759. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Wigg AJ, McCormick R, Wundke R, et al. Efficacy of a chronic disease management model for patients with chronic liver failure. Clin Gastroenterol Hepatol. 2013;11:850–858 e851–854. [DOI] [PubMed] [Google Scholar]
- 29.Runyon BA and Committee APG. Management of adult patients with ascites due to cirrhosis: an update. Hepatology. 2009;49:2087–2107. [DOI] [PubMed] [Google Scholar]
- 30.Singal AG, Rahimi RS, Clark C, et al. An automated model using electronic medical record data identifies patients with cirrhosis at high risk for readmission. Clin Gastroenterol Hepatol. 2013;11:1335–1341 e1331. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Shah R, Haydek C, Mulki R, et al. Incidence and predictors of 30-day readmissions in patients hospitalized with chronic pancreatitis: A nationwide analysis. Pancreatology. 2018;18:386–393. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Figure A1. Timing of readmission for compensated cirrhosis patients. The histogram depicts the percentage of decompensated cirrhosis patients readmitted over a 30 day period with bold arrows marking 50% all 30-day readmission on day 11.
Figure A2. Timing of readmission for decompensated cirrhosis patients. The histogram depicts the percentage of decompensated cirrhosis patients readmitted patients over a 30 day period with bold arrows marking 50% all 30-day readmission on day 12
Figure A3. Length of stay during the index hospitalization. The graph shows the trends in median length of stay for compensated and decompensated cirrhosis patients
Appendix Table 1. ICD-9 & ICD-10 codes for diagnoses and procedures
Appendix Table 2. Baseline characteristics of index patients