Impact of comorbidities on hospital mortality in patients with acute pancreatitis: a population-based study of 110,021 patients

Acute pancreatitis (AP) is a prevalent acute inflammatory disease that affects the pancreas, with an increased incidence in recent years [1, 2]. Most cases are mild with a self-limited course [3]. However, patients with severe acute pancreatitis have a high mortality rate (20–50%) [4‐6]. For this reason, many efforts have been made to find predictors of severity and mortality in patients with AP [7‐11] to identify patients who need admission to an intensive care unit or specific treatment.

In clinical practice, systems such as the Ranson score, the Acute Physiology and Chronic Health Evaluation II (APACHE II) score, the Computed Tomography Severity Index (CTSI), the Bedside Index for Severity in Acute Pancreatitis (BISAP), and various biochemical markers are used to predict severe AP and mortality [3, 12‐16]. However, hospital mortality in AP could also be related to intrinsic patient characteristics, such as individual comorbidities. Most classic scores do not consider comorbidities before admission, except for APACHE II, but are restricted to severe chronic diseases.

According to some previous studies, patients with certain comorbidities, such as obesity [17], hypertriglyceridemia [18], chronic renal failure [19], diabetes [20, 21], and systemic lupus erythematosus [22], are associated with a higher risk of AP severity and mortality. However, few studies currently evaluate the impact of comorbidities on AP severity and mortality.

Our study aimed to determine the relevance of comorbidities and their indexes (Charlson and Elixhauser) as predictors of hospital mortality in patients with AP.

Our study found that pre-admission comorbidities such as heart disease, kidney disease, moderate-severe liver disease, peripheral vascular disease, cerebrovascular disease, and age ≥ 65 years were independently associated with mortality in AP. Charlson and Elixhauser comorbidity indices were independently associated with mortality.

Advanced age has been extensively studied as a marker of severity and mortality in AP. Most studies report longer hospitalization [31, 32] and higher overall mortality from AP in elderly patients [33‐36]. However, other studies have observed that older patients may have a more severe course of AP but do not present increased mortality [37]. Likely explanations explaining advanced age as a risk factor for mortality include a proinflammatory state in older people [38] and increased production of cytokines in elderly patients with sepsis [39]. Other reasons would be delayed diagnosis and treatment due to less clinical and analytical expression [40, 41].

The increase in life expectancy and the aging of the population have been associated with the increase in patients with comorbidities [42, 43], so determining its impact on AP becomes more necessary. The importance of comorbidities in predicting outcomes in other diseases that require acute hospital admission is well known [44‐46]. However, few studies analyze the impact of comorbidities on severity and mortality in AP patients [47, 48]. Additionally, few studies have incorporated comorbidities in their clinical models when evaluating determinants of AP severity. Our study is one of the first studies in the literature that put in relevance the role of comorbidities in AP.

In the last period of our study, we observed an increase in comorbidities and the values of the Charlson and Elixhauser comorbidity indices. These trends could be explained by the increase in life expectancy and the prevalence of chronic diseases in the European population in the last decades [49, 50]. However, despite the increase in comorbidities, hospital mortality has decreased in the period studied. The decrease in mortality is likely due to the advances in critical care medicine, step-up approach to treat infected necrosis, and surgical and endoscopic new approaches [3, 51].

Few studies had previously assessed comorbidity indexes' role in predicting mortality in patients with AP. Previous studies have observed that more comorbidities are associated with organ failure and mortality in patients with AP [47, 52]. In our study, we analyzed the Charlson and Elixhauser indices which are good predictors of mortality in other diseases [53, 54]. Our results showed that values > 1.5 points for both indices are independently associated with hospital mortality in AP after adjusting for age and sex. Future studies that expand knowledge of the effects of comorbidities on complications and mortality in patients with AP will improve the identification of patients at risk and their quality of care.

Regarding other comorbidities, our results align with previous studies stating that pre-existing heart and renal diseases predict mortality in patients with AP [19, 47, 55]. One hypothesis is that intravascular depletion and aggressive fluid resuscitation cause decompensation of previous heart and renal disease [55].

Similarly, to Frey et al. [47], we found an association between liver disease, peripheral vascular disease, and cerebrovascular disease with mortality. However, Murata et al. found no association between these diseases with mortality [55]. The worse results of AP in patients with liver diseases such as cirrhosis could be explained because they present an inflammatory syndrome with arterial vasodilation and release of proinflammatory cytokines that increase the severity of AP [56]. In addition, acute pancreatitis produces significant stress that could decompensate underlying chronic comorbidities and increase the risk of death.

Other major comorbidities before admission, such as chronic lung disease, were not independently associated with mortality, coinciding with the results of Murata et al. [55]. Similar controversial results were found regarding obesity, in which previous studies have observed that obesity is a risk factor for developing local and systemic complications and mortality in patients with AP [17, 57‐59]. We did not find an association between hospital mortality and obesity, but this result has to be taken carefully due to the potential information bias because our results are based on the history of obesity recorded in the medical-administrative database and not on the BMI at hospital admission. In the same line, we did not find a relation between diabetes mellitus and mortality after the multivariable analysis. The literature remains controversial, with reports describing diabetes mellitus as a risk factor for mortality [20, 21]. However, Frey et al. found that diabetes increased the risk of multiple organ failure but was not associated with mortality [47].

Other risk factors classically related to mortality in AP patients failed to represent a risk factor in our population. This was relevant to the role of AP etiology, in which the literature reports controversial results, identifying a more severe course and higher mortality in alcoholic pancreatitis [58, 59]. In contrast, others observed greater severity in biliary pancreatitis [60, 61] or no relation with mortality between both aetiologies [37, 62‐64]. Our study did not observe that acute pancreatitis's biliary or alcoholic etiology was associated with higher hospital mortality. However, in our study, 44.3% of the patients were classified as "unspecified acute pancreatitis,” limiting the precision of our results and constituting a bias regarding the real impact of etiology in AP mortality. In addition, our database does not include other etiologies of acute pancreatitis such as those caused by hypercalcemia, after trauma, viral infections, anatomical variants, iatrogenic after endoscopic retrograde cholangiopancreatography or endoscopic ultrasound-guided interventions [65‐67].

Our study is subject to some limitations. The data analysis from a clinical-administrative base has low level of granularity and does not include some clinical results of interest, such as severity or the evolution of the patient in the medium or long term after their hospital stay. Our study could not identify pancreatic necrosis in the first two years because it began to be considered in the ICD-10 in 2018. Another limitation is the potential underreporting of information because the hospital discharge report may be incomplete or poorly registered by the technical-administrative staff. In our study, we could not identify the etiology of AP in many patients due to a lack of precise coding.

No studies have been published to validate the use of ICD-10 codes for identifying patients with AP using the RAE-CMBD database. However, there are recent studies from Danish [68] and US [69] databases with PPV of 97.3% and 87%, respectively. A recent meta-analysis recommends using ICD codes only in incident cases of AP in adults, where it reaches a PPV of 78% [70]. However, this may be because the studies analyzed used ICD-8, ICD-9, and ICD-10 codes, and the PPV is higher when using the ICD-10 because it is more specific and includes the etiology of pancreatitis [70, 71]. In addition, the studies were carried out in different hospitals in several countries, contributing to the heterogeneity.

We used the primary and the first registered secondary diagnoses to reduce the bias of not including patients with an initial diagnosis of cholelithiasis/choledocholithiasis and AP. A recent study validating ICD codes using primary and secondary diagnoses observed a PPV of 97.3% for AP [68].

The main strength of our study is its large sample size which provides strong statistical power. The RAE-CMBD database is a mandatory registry for the Spanish National Health System, which covers almost 100% of admissions in Spain, reinforcing the external validity of our results. In addition, the database has several internal audit mechanisms and has proven its usefulness for health research [53, 72, 73].