Determinants of in-hospital mortality within 48 hours of admission to the Emergency and Urgent Care Department at University Teaching Hospital, Lusaka, Zambia: a retrospective cross-sectional study

This retrospective study of 385 patients at Zambia's University Teaching Hospital identifies lower Glasgow Coma Scale scores, reduced pulse rates, and out-of-hours (00:00–07:59) admission as independent predictors of in-hospital mortality within 48 hours, highlighting critical areas for improving emergency care protocols in resource-limited settings.

The Gist

Imagine the Emergency Department (ED) at a major hospital as the "front door" to the healthcare system. It's where people rush when they are in immediate danger. This study looked at what happens to patients in the first two days (48 hours) after they walk through that door at the University Teaching Hospital in Lusaka, Zambia.

The researchers wanted to find out: What are the warning signs that a patient might not survive their first two days in the hospital?

Think of the hospital records as a giant puzzle. The researchers took 385 puzzle pieces (patient records) from 2021 and tried to see which pieces fit together to predict a tragedy. They found that out of these 385 patients, nearly half (45.5%) did not make it past the 48-hour mark.

Here is the breakdown of their findings, using simple analogies:

1. The Three "Red Flags"

The study identified three specific things that acted like independent warning lights on a dashboard. If you saw these, the risk of dying within 48 hours went up significantly:

• The "Sleepy" Brain (Glasgow Coma Scale):
  Think of the brain's alertness like a battery charge. The researchers used a score called the Glasgow Coma Scale (GCS) to measure how awake a patient is.
  • The Finding: Patients who arrived with a "low battery" (a lower score, meaning they were very drowsy or unconscious) were much more likely to die. Every little bit of alertness they had was like a shield; the more alert they were, the better their chances.
• The "Slowing Heart" (Pulse Rate):
  This one is a bit tricky. Usually, we think a fast heart is bad. But in this study, the survivors actually had slightly faster heart rates than those who died.
  • The Finding: The patients who died tended to have slower heart rates. The researchers explain this like a car engine: A healthy engine revs up to compensate for trouble. When the engine (the heart) slows down too much, it means the car is running out of fuel and can't compensate anymore. So, a lower pulse was a sign that the body was giving up.
• The "Midnight Shift" (Time of Admission):
  Imagine the hospital as a factory. During the day, there are many managers, tools, and workers. At night, the lights are dimmer, and fewer people are on duty.
  • The Finding: Patients who arrived between midnight and 8:00 AM were much more likely to die than those who arrived during the day. The study suggests that the "night shift" has fewer resources or staff, making it harder to save patients who arrive during those hours.

2. What Didn't Matter as Much

You might think that being older or having a specific type of illness (like a respiratory problem) would be the biggest reason for dying.

• Age: In the initial look, older people seemed to die more often. But when the researchers looked closer at the other factors (like how awake they were or their heart rate), age stopped being a direct cause. It was more like a side effect: older people often arrived with sicker bodies, but it was the sickness (the low alertness or slow heart) that was the real killer, not the number of years they had lived.
• The Unit: More medical patients (like those with infections) died than surgical patients, but again, this was because the medical patients arrived with worse vital signs, not because the "Medical Unit" itself was dangerous.

3. The "Night Shift" Mystery

The finding about the "Midnight Shift" is the most dramatic. The numbers showed that arriving at night made the risk of death jump up massively (over 11 times higher in their math).

• The Caveat: The researchers are very careful to say, "Take this number with a grain of salt." Because so few people arrived during the day in their sample, the math has a wide "margin of error." However, the direction is clear: arriving at night is risky, likely because the hospital is quieter and staffed differently at that time.

The Bottom Line

The study concludes that to save more lives in the first 48 hours, the hospital needs to focus on three things:

1. Check the "Battery": Always check how awake a patient is (GCS score) immediately.
2. Watch the Engine: Pay close attention to the pulse rate; a slowing heart is a sign of trouble.
3. Staff the Night Shift: Make sure there are enough doctors and nurses working between midnight and 8:00 AM to handle emergencies effectively.

The researchers emphasize that these are the specific factors they found in their data. They suggest that by fixing these specific operational and monitoring issues, the hospital could potentially prevent many of these early deaths.

Technical Summary

Technical Summary: Determinants of 48-Hour In-Hospital Mortality in a Zambian Emergency Department

Problem Statement
In-hospital mortality occurring within 48 hours of admission to an Emergency Department (ED) serves as a critical sentinel indicator of acute care failures, particularly in low- and middle-income countries (LMICs). While effective emergency systems could prevent up to 50% of deaths in these settings, early mortality in sub-Saharan Africa (SSA) is exacerbated by structural barriers such as staffing shortages, diagnostic limitations, and delayed referrals. Despite Zambia's National Health Strategic Plan targeting a reduction in 48-hour mortality to 3 per 1,000 admissions, the specific determinants of early mortality at the University Teaching Hospital (UTH)—the nation's sole quaternary referral center—remain poorly characterized. Existing data are largely limited to specific disease cohorts (e.g., sepsis, COVID-19) rather than a comprehensive analysis across the full spectrum of emergency presentations.

Methodology
This study employed a retrospective cross-sectional design utilizing patient records from the Emergency and Urgent Care Department (EUCD) at UTH, Lusaka, covering the period from January 1 to December 31, 2021.

• Data Source and Sampling: Data were extracted from the District Health Information System 2 (DHIS2). A simple random sampling method was used to select a sample size of 385 patients, calculated based on a conservative prevalence estimate of 50%, a 95% confidence level, and a 5% margin of error.
• Inclusion/Exclusion: Eligible patients were those admitted to the EUCD in 2021 with documented outcomes (death within 48 hours or survival beyond 48 hours). Records with >50% missing key variables were excluded.
• Variables: The primary outcome was in-hospital mortality within 48 hours. Explanatory variables included demographics (age, sex), clinical factors (vital signs, Glasgow Coma Scale [GCS], admission reason, comorbidities), and hospital factors (time/day of admission, unit type, referring facility level).
• Statistical Analysis: Descriptive statistics were generated, followed by univariate analyses (Mann–Whitney U and Chi-squared tests). Variables with $p < 0.20$ in univariate analysis were retained for multivariable logistic regression. Multicollinearity was assessed via Variance Inflation Factors (VIF < 5). Model fit was evaluated using the Hosmer–Lemeshow test and Receiver Operating Characteristic (ROC) analysis. Analyses were conducted using STATA 16.1 MP.

Key Results
Of the 385 patients analyzed, 175 (45.5%) died within 48 hours of admission.

• Demographics: Patients who died were significantly older (median age 45 vs. 37.5 years, $p < 0.001$) than survivors. Sex was not significantly associated with mortality.
• Univariate Findings: Significant associations with mortality were found for age, respiratory admission reason, oxygen saturation, GCS score, and admission to the medical unit versus the surgical unit.
• Multivariable Analysis: Three variables were independently associated with 48-hour mortality in the adjusted model:
  1. Pulse Rate: Inversely associated with mortality (Adjusted Odds Ratio [aOR] = 0.98; 95% CI: 0.95–1.00; $p = 0.036$). Each unit increase in pulse rate was associated with a 2% reduction in the odds of death, interpreted as a sign of compensatory physiological response in survivors.
  2. Glasgow Coma Scale (GCS): Higher admission GCS scores were protective (aOR = 0.75; 95% CI: 0.63–0.90; $p = 0.002$). Each one-point increase in GCS reduced the odds of death by 25%.
  3. Time of Admission: Admissions between 00:00 and 07:59 were associated with substantially higher mortality compared to 08:00–23:59 (aOR = 11.44; 95% CI: 1.19–109.96; $p = 0.035$). This effect emerged only after adjustment, suggesting case-mix confounding masked the association in univariate analysis.
• Model Performance: The model demonstrated good discriminatory ability with an Area Under the Curve (AUC) of 0.81 and adequate goodness-of-fit (Hosmer–Lemeshow $p = 0.45$).

Significance and Claims
The authors position this study as one of the first to systematically characterize 48-hour mortality determinants across the full spectrum of emergency admissions at a Zambian quaternary center. The findings underscore the critical role of modifiable clinical and operational factors over non-modifiable demographics in explaining early mortality.

The paper claims immediate relevance for clinical practice and health policy in resource-constrained settings:

1. Triage Protocols: The strong association between low GCS and mortality supports the standardization of systematic GCS assessment as a core triage criterion.
2. Staffing and Operations: The significant risk associated with overnight admissions (00:00–07:59) highlights the need for targeted staffing improvements and supervising clinician availability during these hours to mitigate structural vulnerabilities.
3. Risk Stratification: The study advocates for the evaluation and validation of composite vital sign monitoring tools, such as the National Early Warning Score (NEWS), to improve real-time risk stratification.

Limitations and Caution
The authors maintain a modest tone regarding the study's generalizability and specific findings. They acknowledge that the retrospective design relies on the completeness of routine records, noting significant missing data for GCS (38.2%) and temperature (56.4%). The study is restricted to a single tertiary facility, limiting direct applicability to lower-level health centers. Furthermore, the authors explicitly caution that the wide confidence interval for the "out-of-hours" finding (1.19–109.96) reflects sparse data in the reference category and warrants cautious interpretation and validation in future prospective studies. They also note that process variables, such as nurse-to-patient ratios and time-to-intervention, were unavailable for analysis.