Cirrhosis is the common histological end point of chronic liver injury, characterised by diffuse fibrosis and conversion of normal hepatic architecture into regenerative nodules [1]. Its clinical course is conventionally divided into a compensated phase, which may last many years, and a decompensated phase heralded by variceal bleeding, ascites, jaundice or hepatic encephalopathy. Hepatic encephalopathy (HE) — a reversible neuropsychiatric syndrome caused by accumulation of gut-derived neurotoxins that the failing liver can no longer clear — is among the most ominous of these events [2,3].

The prognostic weight of a first episode of overt encephalopathy is considerable. Bustamante et al. demonstrated that survival falls sharply once encephalopathy appears, with roughly 42% of patients alive at one year and 23% at three years [4]. Cordoba et al., analysing a large European cohort of cirrhotic patients hospitalised with HE, reported markedly higher mortality where encephalopathy occurred in the context of acute-on-chronic liver failure [5]. The syndrome is thus simultaneously a marker of hepatic reserve exhausted and, because it is potentially reversible, an opportunity for intervention.

That opportunity turns on the precipitating factor. Most episodes in cirrhosis follow an identifiable trigger — gastrointestinal haemorrhage, infection, constipation, electrolyte disturbance or sedative use — rather than spontaneous decompensation alone [5,6]. These triggers differ substantially in their intrinsic lethality. Variceal bleeding carries a reported mortality of 20–30% at the first episode independent of its encephalopathic consequences [7], while bacterial infection in cirrhosis provokes a systemic inflammatory response that may cascade into multi-organ failure [8,9]. By contrast, constipation and mild electrolyte derangement are usually corrected within days. It follows that outcome after an episode of HE should depend not only on the depth of hepatic decompensation but also on which event precipitated it — yet stratified outcome data by precipitant remain relatively sparse, particularly from South Asian centres.

Quantifying underlying hepatic reserve is central to interpreting these outcomes. The Child–Pugh score, derived by Pugh et al. from serum bilirubin, albumin, prothrombin time, ascites and encephalopathy grade, remains a robust bedside index of severity [10]. The Model for End-Stage Liver Disease (MELD), developed by Kamath et al. from bilirubin, creatinine and the international normalised ratio, provides a continuous and objective estimate of three-month mortality and underpins transplant allocation in most programmes [11]. Because hyponatraemia independently predicts death in cirrhosis, Biggins et al. incorporated serum sodium to yield the MELD-Na score, which improves discrimination further [12]. Applying all three to a cohort defined by the presence of encephalopathy allows the depth of decompensation at presentation to be characterised precisely.

Length of hospital stay is a third and complementary outcome, capturing both clinical burden and health-system cost. In a district hospital serving a largely rural and economically vulnerable population, prolonged admission carries substantial indirect costs for families, and identifying which precipitants drive longer stays has direct planning value. The present study was therefore undertaken to determine in-hospital mortality stratified by precipitating factor, to characterise disease severity using the Child–Pugh, MELD and MELD-Na scores, and to compare length of stay across the major precipitants of hepatic encephalopathy in cirrhosis.

This prospective, descriptive, hospital-based observational study was conducted in the Department of General Medicine, Ballari Medical College and Research Centre (formerly Vijayanagar Institute of Medical Sciences), Ballari, Karnataka, a tertiary care teaching hospital serving a predominantly rural district population. Ethical considerations: Institutional Ethics Committee approval was obtained prior to commencement. Written informed consent was taken from each participant, or from a legally acceptable representative where the depth of encephalopathy precluded personal consent. Study population and eligibility: All patients admitted with hepatic encephalopathy during the study period were screened consecutively. Eligible patients were aged above 12 years, of either sex, with established cirrhosis of any aetiology and hepatic encephalopathy of any grade, including minimal hepatic encephalopathy. Exclusions were known psychiatric disorder or ongoing psychiatric treatment, altered sensorium attributable to metabolic disease or head injury, and acute alcoholic intoxication or alcohol withdrawal, each of which confounds the clinical diagnosis. Sample size estimation: Sample size was calculated using nMaster software version 2.0 with the single-proportion formula, taking the reference proportion from Poudyal et al. [13]. With an expected proportion of 0.333, a two-sided alpha of 0.05 and a precision of 10% at 95% confidence, the required sample was 85 subjects. Data collection and clinical assessment: A structured questionnaire captured present and past illness, with directed enquiry into haematemesis, melaena, constipation, vomiting, diarrhoea, oliguria, fever, bleeding manifestations, dietary protein intake, recent paracentesis, trauma and surgery. Personal history of alcohol, tobacco and intravenous drug use, and use of sedatives, diuretics, tranquillisers, analgesics and cough preparations were documented. All patients were examined for jaundice, pallor, fever, asterixis, hydration, pedal oedema and ascites, with full abdominal and neurological examination. Encephalopathy was graded clinically from 1 to 4. Investigations: Complete blood count, liver and renal function tests, random blood sugar, serum electrolytes, serum albumin and coagulation profile were performed in every patient. Abdominal ultrasonography assessed liver and splenic size, parenchymal echogenicity, portal vein diameter and ascites; wherever ascites was detected, a diagnostic tap was carried out to exclude spontaneous bacterial peritonitis. Severity scoring: The Child–Pugh score was computed for each patient from serum bilirubin, serum albumin, prothrombin time, ascites and encephalopathy grade [10]. The MELD score was calculated from serum bilirubin, serum creatinine and the international normalised ratio [11], and the MELD-Na score derived by incorporating serum sodium into the MELD equation [12]. Outcome measures: The primary outcome was in-hospital mortality, recorded as a binary variable (died / survived to discharge) and analysed against the precipitating factor assigned at admission. Secondary outcomes were the mean Child–Pugh, MELD and MELD-Na scores for the cohort, and the mean duration of hospital stay in days, computed separately for each of the four major precipitant categories. Case fatality rate for each precipitant was derived as the number of deaths divided by the number of patients in whom that factor was identified. Statistical analysis: Data were entered and analysed in IBM SPSS Statistics version 23.0. Categorical variables were expressed as frequencies and percentages and continuous variables as mean and standard deviation. Associations between categorical variables, including the relationship between precipitating factor and mortality, were tested using the chi-square test or Fisher's exact test where expected cell counts were small. A probability value below 0.05 was considered statistically significant.

Eighty-five cirrhotic patients admitted with hepatic encephalopathy were followed to discharge or death.

Table 1. Baseline characteristics of the cohort (n = 85)

The cohort was young, male and predominantly alcohol-related, with more than two-thirds of patients aged between 31 and 50 years. Overall in-hospital mortality was 34.1%, meaning that approximately one in every three patients admitted with hepatic encephalopathy did not survive to discharge — a figure that establishes the clinical gravity of the syndrome in this setting and provides the denominator against which the factor-specific analyses below should be read.

Table 2. Mean liver disease severity indices (n = 85)

The mean Child–Pugh score of 10.29 places the average patient firmly in Child class C, the category denoting advanced hepatic decompensation and poor prognosis. The mean MELD score of 24.06 corresponds to a substantial predicted three-month mortality and, in a transplant-capable system, would place these patients at high priority for listing.

That the mean MELD-Na score of 25.38 exceeds the MELD score by 1.32 points indicates that hyponatraemia was prevalent across the cohort and was adding measurably to prognostic risk beyond that captured by bilirubin, creatinine and INR alone. Taken together, the three indices confirm that patients presented not with early decompensation but with severely exhausted hepatic reserve.

Table 3. In-hospital mortality by individual precipitating factor (n = 85)

Case fatality differed markedly across precipitants. Pneumonia carried the highest proportional risk, with 3 of 5 patients dying (60.0%), although the small numerator warrants caution. Gastrointestinal bleeding was close behind at 54.5% and, being far commoner, dominated the absolute mortality burden: 18 of the 29 deaths in the study (62.1%) occurred in patients precipitated by haemorrhage. Constipation (26.3%) and electrolyte imbalance (22.2%) occupied an intermediate position. Notably, spontaneous bacterial peritonitis was associated with a case fatality of only 10.0% — considerably lower than pneumonia — most plausibly because routine diagnostic paracentesis in all patients with ascites permitted early detection and prompt antibiotic therapy. No deaths occurred among the 9 patients whose episodes were precipitated by overdiuresis, paracentesis or urinary tract infection, indicating that these are readily reversible triggers when the underlying trigger is removed.

Table 4. In-hospital mortality by grouped precipitant category (n = 85)

Aggregating the individual factors clarifies the clinical priority. Gastrointestinal bleeding stands apart: it was both the commonest precipitant and by far the most lethal, generating almost two-thirds of all deaths in the study. Constipation, despite a moderate case fatality of 26.3%, contributed a further 17.2% of deaths by virtue of its frequency. The aggregate infection category showed a case fatality of 22.2%, but this figure conceals the wide divergence between pneumonia (60.0%) and the other infective entities. The complete absence of mortality in the iatrogenic group confirms that procedure- and diuretic-related encephalopathy, while entirely preventable, is not in itself life-threatening once recognised.

Table 5. Mean duration of hospital stay by precipitating factor

Mean length of stay clustered narrowly between 9.7 and 10.6 days across all four major categories, a spread of less than a single day. Constipation was associated with the longest admission at 10.57 days, exceeding gastrointestinal bleeding by 0.68 days despite its far lower mortality. This apparent paradox is readily explained: patients with variceal haemorrhage either respond to endoscopic and haemodynamic intervention and are discharged, or die early, so that deaths truncate the observed stay, whereas constipation-related encephalopathy typically resolves slowly with lactulose titration in patients who survive. Length of stay in this cohort should therefore be interpreted as a measure of resource utilisation rather than as a proxy for illness severity.

Table 6. Combined outcome summary by precipitant category

This synthesis integrates frequency, lethality and resource use into a single view. Gastrointestinal bleeding emerges unambiguously as the priority target, combining the highest frequency with the highest category-level case fatality. Constipation ranks second on frequency and first on length of stay, making it the leading driver of bed occupancy. Infection occupies a middle position overall but is internally heterogeneous, with pneumonia behaving quite differently from spontaneous bacterial peritonitis. The iatrogenic group, though small and non-fatal, represents wholly avoidable admissions.

The overall in-hospital mortality of 34.1% in this cohort confirms that hepatic encephalopathy in cirrhosis remains a lethal complication despite its theoretical reversibility. This figure is consistent with the poor survival following a first encephalopathic episode described by Bustamante et al., who reported one-year survival of approximately 42% after the onset of encephalopathy [4], and with the substantial in-hospital mortality documented by Cordoba et al. in cirrhotic patients hospitalised with HE, in whom outcome depended heavily on the presence of concurrent organ failure [5]. Gastrointestinal bleeding was both the most frequent precipitant and the dominant contributor to mortality, accounting for 18 of 29 deaths with a case fatality of 54.5%. This accords with D'Amico et al., who reported a 20–30% mortality attending the first episode of variceal haemorrhage [7], and the higher figure here is unsurprising given that our patients had, by definition, already progressed to encephalopathy at the time of bleeding. Haemorrhage acts through two simultaneous mechanisms — a large intraluminal nitrogen load and hypovolaemic renal hypoperfusion impairing ammonia clearance [14] — and it is the additive effect of these that makes it so dangerous. The implication is that early endoscopic haemostasis, adequate volume resuscitation and prophylactic antibiotics constitute the interventions with the greatest potential yield in this population. Pneumonia carried the highest proportional mortality at 60.0%, consistent with the observation of Blei and Córdoba that severe infection in cirrhosis is associated with high fatality [15]. The contrast with spontaneous bacterial peritonitis, at 10.0%, is instructive. Since diagnostic paracentesis was performed routinely in every patient with ascites, SBP was likely detected early and treated promptly, whereas pneumonia in an encephalopathic patient may present late and is compounded by aspiration risk. This divergence argues against treating "infection" as a single prognostic category, as several earlier series have done [8,16]. The mean Child–Pugh score of 10.29 places the cohort in class C, and the mean MELD of 24.06 corresponds to high predicted short-term mortality, validating the prognostic frameworks of Pugh et al. and Kamath et al. respectively [10,11]. That the mean MELD-Na exceeded MELD by 1.32 points reflects widespread hyponatraemia and supports the finding of Biggins et al. that incorporating serum sodium improves mortality prediction in cirrhosis [12]. These scores indicate that patients reached hospital only at an advanced stage of decompensation, a pattern characteristic of resource-limited settings where earlier presentation is uncommon. The finding that constipation was associated with the longest mean stay (10.57 days) while carrying only intermediate mortality echoes the observation of Riggio et al. linking constipation to prolonged hospitalisation and greater encephalopathy severity [17]. Stay durations for the remaining categories were closely comparable, in keeping with the hospitalisation patterns reported by Bustamante et al. and Ginès et al. [4,16]. Limitations: The single-centre design and sample of 85 restrict generalisability, and several precipitant subgroups contained fewer than ten patients, so that case fatality estimates for pneumonia and the iatrogenic group carry wide uncertainty. Assignment of a single precipitant may misattribute outcome where multiple triggers coexisted. Mortality was censored at discharge, so post-discharge deaths were not captured, and no adjustment was made for confounding by baseline MELD across precipitant groups.

Among 85 cirrhotic patients admitted with hepatic encephalopathy, in-hospital mortality was 34.1%, confirming that one in three such admissions ends in death despite the syndrome's potential reversibility. Outcome varied substantially with the precipitating event. Pneumonia carried the highest case fatality (60.0%) and gastrointestinal bleeding the greatest absolute burden, generating 62.1% of all deaths at a case fatality of 54.5%. Constipation (26.3%) and electrolyte imbalance (22.2%) were of intermediate severity, spontaneous bacterial peritonitis was comparatively survivable at 10.0%, and no deaths occurred in the iatrogenic or procedure-related group. Mean Child–Pugh (10.29), MELD (24.06) and MELD-Na (25.38) scores establish that patients presented with advanced Child class C disease and high predicted short-term mortality, with hyponatraemia adding further prognostic weight. Mean hospital stay was longest for constipation (10.57 days) and broadly similar across the other categories.

These findings define a clear clinical priority. Rapid endoscopic control of variceal haemorrhage with volume resuscitation and prophylactic antibiotics, together with early recognition and aggressive treatment of pneumonia, address the two precipitants responsible for the great majority of deaths. Routine diagnostic paracentesis appears to have rendered spontaneous bacterial peritonitis a survivable trigger and should be maintained as standard practice. Given the advanced MELD-Na scores at presentation, earlier identification and referral of decompensating cirrhotic patients, and access to transplantation where feasible, remain essential to improving survival. Larger multicentre studies with post-discharge follow-up and adjustment for baseline severity are required to confirm these factor-specific outcome differences.