External validation and comparison of SAFE QUAKE and dialysis score in earthquake related crush injuries

Abstract

Background

This study aimed to compare the predictive accuracy of two scoring systems—Seismic Assessment of Kidney Function to Rule Out Dialysis Requirement (SAFE-QUAKE) and the Dialysis Score—developed to assess the need for dialysis in patients with crush injuries admitted to the emergency department (ED) following the February 6 Türkiye earthquakes.

Methods

In this retrospective observational study, the performance of the SAFE-QUAKE and Dialysis Score systems were evaluated using data from a university hospital that was independent from the centers where these scores were originally developed. The study included patients admitted to the ED after the same earthquakes.

Results

The SAFE-QUAKE score demonstrated a higher negative predictive value (NPV) for ruling out dialysis (93.4%), whereas the Dialysis Score had superior positive predictive value (PPV) (71.0%). Although the area under the receiver operating characteristic curve (AUROC) was higher for SAFE-QUAKE [0.894 (95% CI: 0.831–0.957)] than for the Dialysis Score [0.836 (95% CI: 0.738–0.934)], this difference was not statistically significant (Z = 1.415; p = 0.157). According to the Net Reclassification Index (NRI), SAFE-QUAKE provided a clear advantage in correctly reclassifying non-dialysis patients to a lower-risk category [NRI⁻ = 0.491 (95% CI: 0.321–0.643)], correctly downgrading 74.5% of such patients while misclassifying only 25.5%. Across various hypothetical prevalence rates, SAFE-QUAKE consistently provided higher NPVs (e.g., 98.9% at 5% prevalence), while the Dialysis Score maintained higher PPVs (e.g., 28% at 5% prevalence; 88.0% at 50% prevalence).

Conclusion

Following catastrophic disasters such as the February 6 Türkiye earthquakes, the capacity of healthcare facilities to provide dialysis becomes a key factor in ED decision-making. When patients are to be transferred to facilities without dialysis capabilities, triage based on the SAFE-QUAKE score can help safely identify those at low risk for dialysis. Conversely, if the receiving center has dialysis capacity, the Dialysis Score can be used to prioritize patients more likely to require dialysis. The sequential use of both scores may enhance triage accuracy by prioritizing SAFE-QUAKE for rapid ED assessment and Dialysis Score for refining interfacility transfer decisions based on resource intensity.

Clinical trial number

Not applicable.

Introduction

According to the Emergency Events Database, a total of 399 natural hazard-related disasters occurred in 2023, resulting in 86,473 fatalities and affecting 93.1 million people worldwide [1]. Of these deaths, approximately 53,000 were caused by the catastrophic disaster that struck on February 6, severely affecting 11 provinces in Türkiye and causing widespread destruction and loss of life in war-torn Syria as well [2, 3].

On February 6, 2023, Türkiye experienced one of the most devastating disasters in its history. The enormity of the destruction was primarily due to two high-magnitude earthquakes—measuring 7.7 and 7.6 on the Richter scale—that occurred just nine hours apart [4]. The first earthquake lasted for 80 s and the second for 45 s, both being intensely felt across the region. As a result, hundreds of thousands of buildings collapsed. Thousands of people lost their lives inside these structures, and many more were trapped or crushed beneath the rubble.

Following the earthquake, a large-scale search and rescue operation was launched. However, many hospitals and emergency departments (EDs) were also damaged or destroyed by the disaster [5]. Although field hospitals could not be established immediately, they were eventually set up during the later stages of the rescue efforts with both national and international support. Prior to the establishment of these field hospitals, the disaster area was characterized by: (1) collapsed buildings with thousands of injured individuals, (2) damaged hospitals, and (3) healthcare workers and patients unwilling to enter hospital buildings due to ongoing aftershocks. Consequently, only EDs remained functional, and the management of this catastrophic disaster in many of the affected provinces had to be carried out through these departments [2].

According to official statements, 53,537 people lost their lives and 107,213 were injured as a result of the earthquake [6]. The high number of collapsed buildings significantly contributed to the large number of entrapment-related deaths. Moreover, the rescue of an overwhelming number of trapped individuals further complicated patient management both at the scene and in EDs. In response, Türkiye’s health authorities implemented a phased evacuation strategy. Initially, injured patients were transported by land, sea, and air to nearby provinces with less structural damage (such as Adana, Mersin, and Diyarbakır). Later, patients were transferred by air to more distant cities, including Ankara, Istanbul, and Izmir [7].

Catastrophic disasters cause unique impacts in the affected areas due to their magnitude, and the lessons learned from such events offer opportunities to question and improve global disaster management practices [8]. Koenig and colleagues previously examined earthquakes among disasters capable of causing catastrophic destruction [9]. In 1996, they identified disaster triage as the most strategic point in the medical management of victims during such events and proposed the use of Simple Triage and Rapid Treatment (START) along with the Secondary Assessment of Victim Endpoint (SAVE) system for managing casualties in catastrophic scenarios. The SAVE system includes triage scores based on clinical parameters such as the Glasgow Coma Scale, Mangled Extremity Severity Score, and Burn Triage Score, enabling evaluation based on the patient’s condition.

During the February 6 earthquakes, the emergency response infrastructure—including search and rescue operations, hospitals, and prehospital systems—was significantly constrained due to the overwhelming number of casualties and the damage caused by the earthquake. As a result, standard triage protocols could not be implemented in the field. Many injured individuals were brought directly to EDs by themselves or by their relatives. In this context, the triage systems recommended by Koenig et al., such as START and SAVE, could not be applied in the disaster field, as the scale of destruction affected prehospital care and search-and-rescue (SAR) teams as well. This led to the EDs becoming the de facto frontline of disaster response [10].

Following this experience, the Disaster Committee of the Emergency Medicine Association of Türkiye (EMAT) conducted a study evaluating the use of a modified SAVE algorithm—integrated with START and a dialysis scoring system—not in the field, but within EDs [11]. One of these scoring systems, SAFE-QUAKE (Seismic Assessment of Kidney Function to Rule Out Dialysis Requirement), was developed in Diyarbakır, a relatively less-affected neighboring province. SAFE-QUAKE is a predictive scoring system that can be quickly and easily calculated in the ED to assess dialysis requirements in patients with earthquake-related injuries [12]. Despite its limited validation, SAFE-QUAKE was proposed as a potentially useful tool in the ED setting for identifying patients at risk of requiring dialysis.

Following the same devastating earthquake, emergency medicine researchers also observed the challenges of managing patients with crush injuries in emergency settings. This led to the development of another predictive model, the Dialysis Score, in a subsequent study conducted in Ankara—one of the referral cities that received earthquake victims. The Dialysis Score was designed to estimate the need for dialysis in patients with crush-related injuries [13].

We hypothesized that SAFE-QUAKE and the Dialysis Score—two recently developed, earthquake-specific tools for predicting dialysis requirement—would demonstrate complementary predictive strengths when externally validated in a setting independent from their original development centers. These two scores were chosen because they are, to our knowledge, the only validated scoring systems developed specifically in the context of the February 6, 2023 Türkiye earthquakes to aid ED triage decisions for crush injury patients at risk of acute kidney injury requiring dialysis.

Methods

Study design and setting

This study is a single-center, retrospective, observational investigation conducted to compare the diagnostic performance of two scoring systems—SAFE-QUAKE (developed in Diyarbakır, a province affected by seismic shaking, where buildings collapsed and patients presented both locally and via intercity road transport) and the Dialysis Score (developed in Ankara, a non-seismically affected province, geographically distant from the disaster zone, where patients arrived either on foot or via air medical evacuation)—in predicting dialysis requirements among patients admitted to the ED with crush injuries following the February 6, 2023 Kahramanmaraş earthquakes.

The present study was carried out in the ED of a tertiary university hospital located in Mersin—a province that experienced seismic shaking but no structural building collapse. This city received patients from other affected provinces through outpatient presentations, as well as road, air, and sea transfers.

Since the early days of the earthquake, the study center functioned as a tertiary care referral hospital, accepting injured patients from 10 other earthquake-affected provinces and playing a key role in the post-disaster healthcare response. Due to these characteristics, the hospital provided an appropriate setting and patient population to evaluate and compare the performance of both scoring systems independently from the centers where they were originally developed.

Ethical approval was obtained from the local ethics committee prior to study initiation (Approval Date: 30.04.2025, Approval No: 452). All procedures were conducted in accordance with the principles of the Declaration of Helsinki. Due to the retrospective nature of the study and the use of anonymized patient data, the requirement for informed consent was waived by the ethics committee.

Participant selection

Patients admitted to the ED with crush injuries resulting from the February 6, 2023 Kahramanmaraş earthquakes were retrospectively screened. Inclusion criteria for the study were as follows: Aage ≥ 18 years; presentation to the ED following the earthquake with documented evidence of injury due to structural collapse or prolonged entrapment under debris; presence of clinical signs consistent with crush injury—such as localized swelling, severe pain in a compressed extremity, skin discoloration, diminished peripheral pulses, or signs of compartment syndrome; availability of sufficient clinical and laboratory data to calculate both the SAFE-QUAKE and Dialysis Scores; and clearly documented information regarding whether dialysis was performed during hospitalization. Patients were excluded if they were transferred to another healthcare facility during follow-up, had incomplete medical records, or if their trauma was unrelated to the earthquake.

Data collection

Patient data were retrospectively reviewed through the hospital information management system and patient discharge records. The data were obtained from documents recorded during the patients’ initial ED visits, including diagnosis, vital signs, laboratory test results, imaging findings, and follow-up records. The collected data were classified and analyzed under five main categories: (1) Demographic data included the patients’ age (years) and sex; (2) Biochemical and hematological parameters included serum sodium (mmol/L), potassium (mmol/L), chloride (mmol/L), bicarbonate (mmol/L), pH (unitless), pCO₂ (mmHg), blood urea nitrogen [BUN] (mg/dL), creatinine (mg/dL), albumin (g/dL), calcium (mg/dL), and C-reactive protein [CRP] (mg/L), as well as hemoglobin (g/dL), hematocrit (%), leukocyte (×10³/µL), neutrophil (×10³/µL), lymphocyte (×10³/µL), neutrophil-to-lymphocyte ratio [NLR] (unitless), platelet-to-lymphocyte ratio [PLR] (unitless), platelet count (×10³/µL), and mean platelet volume [MPV] (fL); in addition, markers of muscle and tissue damage, including creatine kinase [CK] (U/L), AST (U/L), ALT (U/L), and LDH (U/L), were evaluated; (3) Clinical trauma-related interventions included the number of trauma locations (count), presence of compartment syndrome, performance of fasciotomy, and amputation status; (4) Scoring calculations were performed for the SAFE-QUAKE score (entrapment duration [hours], pH, creatinine [mg/dL], LDH [U/L], AST/ALT ratio [unitless]) and the Dialysis Score (number of traumatized extremities [count], albumin [g/dL], CK [U/L]); (5) The outcome measure was whether the patient received dialysis (renal replacement therapy) during hospitalization, confirmed through discharge notes and hemodialysis treatment forms.

To ensure data reliability, all data were independently entered by two researchers and reviewed by a third researcher through random sample validation. In cases of missing or conflicting records, patient files were re-examined and the most accurate data was retrieved whenever possible.

Scores calculation

In this study, two different scoring systems developed to predict the need for dialysis in earthquake-related trauma patients were utilized: the SAFE-QUAKE score and the Dialysis Score. The SAFE-QUAKE score was developed based on significant differences observed in laboratory parameters, including duration of entrapment, pH, serum creatinine, LDH, and AST/ALT ratio [12]. In this scoring system, one point was assigned for each of the following criteria: entrapment duration < 45 h, pH > 7.31, creatinine < 2 mg/dL, LDH < 1600 U/L, and AST/ALT ratio < 2.4, yielding a total score ranging from 0 to 5.

The Dialysis Score, on the other hand, was developed in a larger cohort study to predict the need for renal replacement therapy (RRT) due to trauma-induced acute kidney injury [13]. This scoring system includes three variables found to be significantly associated with RRT in multivariate analysis: the number of traumatized body regions (sides), serum albumin level (g/dL), and CK level (U/L). The score increases with the number of injured sides, decreases in albumin level, and rises in CK level. Each variable contributes proportionally to the score, resulting in a total ranging from 0 to 7.

Both scores were retrospectively calculated based on the initial admission data of the patients and used in subsequent outcome analyses.

Analysis

All analyses were performed using R version 4.4.2 (R Foundation for Statistical Computing, Vienna, Austria). Descriptive statistics were expressed as mean ± standard deviation (SD) for normally distributed continuous variables, and as median with interquartile range [IQR] for non-normally distributed variables, based on visual inspection of histograms. Categorical variables were summarized as counts and percentages (n [%]). Between-group comparisons were conducted using Student’s t-test or the Mann–Whitney U test for continuous variables, and the Chi-square or Fisher’s exact test for categorical variables, as appropriate. Receiver operating characteristic (ROC) analysis was used to assess the discriminatory performance of both scores for predicting dialysis requirement. Area under the ROC curve (AUROC) with 95% confidence intervals (CI) was calculated. Optimal cutoff values for each score in predicting dialysis requirement were derived from the study population using the Youden Index, rather than predefined thresholds, to ensure the best balance between sensitivity and specificity. Sensitivity, specificity, positive likelihood ratio (+ LR), and negative likelihood ratio (–LR), all with 95% CIs, were reported at the optimal thresholds. Comparisons between AUROCs were conducted using DeLong’s test for paired curves. Calibration was assessed via graphical calibration plots. Reclassification performance was evaluated using the category-free Net Reclassification Index (NRI), along with separate NRI⁺ and NRI⁻ components, with 95% CIs. Additionally, we calculated prevalence-weighted positive predictive values (PPVs) and negative predictive values (NPVs) for each score across a range of hypothetical prevalence values (5–50%) to illustrate real-world applicability. These were plotted to visualize predictive trade-offs. All statistical tests were two-sided, and a p-value less than 0.05 was considered indicative of statistical significance.

Results

A total of 163 patients were included, of whom 28 (17.2%) required dialysis. Table 1 summarizes baseline demographics, laboratory values, and scores stratified by dialysis requirement. The dialysis group had lower mean age [37.46 ± 15.05 vs. 45.16 ± 16.84 years; Δ − 7.70 (95% CI − 14.14 to − 1.26), p = 0.020] and lower serum sodium [130.06 ± 5.94 vs. 134.99 ± 5.63 mmol/L; Δ − 4.94 (–7.41 to − 2.46), p < 0.001], bicarbonate [16.80 ± 3.85 vs. 22.91 ± 4.27 mmol/L; Δ − 6.11 (–7.76 to − 4.47), p < 0.001], and albumin [22.84 ± 4.06 vs. 27.88 ± 4.27 g/dL; Δ − 5.03 (–6.75 to − 3.31), p < 0.001]. Serum potassium was substantially higher in dialysis patients [5.75 ± 0.81 vs. 4.14 ± 0.57 mmol/L; Δ 1.61 (1.28 to 1.93), p < 0.001], as were creatinine [3.92 vs. 0.80 mg/dL, p < 0.001] and blood urea nitrogen levels [54.91 vs. 21.50 mg/dL, p < 0.001]. Median SAFE-QUAKE and Dialysis Scores were significantly different between groups [2.5 vs. 5 and 3 vs. 0, respectively; both p < 0.001].

Table 1.

Demographics, laboratory parameters, and scores according to Dialysis requirement

Variable	No Dialysis (n = 135)	Dialysis (n = 28)	Δ Mean (95% CI) / Median [IQR]	p
Age, years	45.16 ± 16.84	37.46 ± 15.05	–7.70 (–14.14 to − 1.26)	0.020
Sex, female (%)	68 (50.4%)	17 (60.7%)	–	0.430
Potassium, mmol/L	4.14 ± 0.57	5.75 ± 0.81	1.61 (1.28 to 1.93)	< 0.001
Sodium, mmol/L	134.99 ± 5.63	130.06 ± 5.94	–4.94 (–7.41 to − 2.46)	< 0.001
Chloride, mmol/L	106.95 ± 4.40	102.49 ± 6.66	–4.46 (–7.14 to − 1.78)	0.002
Bicarbonate, mmol/L	22.91 ± 4.27	16.80 ± 3.85	–6.11 (–7.76 to − 4.47)	< 0.001
pCO₂, mmHg	40.30 ± 7.07	34.77 ± 7.92	–5.53 (–8.80 to − 2.25)	0.002
pH	7.38 [7.34–7.41]	7.30 [7.24–7.35]	–	< 0.001
BUN, mg/dL	21.50 [14.95–32.01]	54.91 [40.65–88.79]	–	< 0.001
Creatinine, mg/dL	0.80 [0.58–1.04]	3.92 [2.54–4.82]	–	< 0.001
Albumin, g/dL	27.88 ± 4.27	22.84 ± 4.06	–5.03 (–6.75 to − 3.31)	< 0.001
Calcium, mg/dL	8.43 ± 0.84	7.31 ± 0.96	–1.12 (–1.52 to − 0.73)	< 0.001
C-reactive protein, mg/L	116.10 ± 70.17	159.04 ± 73.59	42.94 (12.24 to 73.64)	0.007
SAFE-QUAKE score [median, IQR]	5 [4–5]	2.5 [2–3]	–	< 0.001
Dialysis score [median, IQR]	0 [0–1]	3 [1–4]	–	< 0.001

BUN, blood urea nitrogen; IQR, interquartile range; SD, standard deviation; CI, confidence interval; pCO₂, partial pressure of carbon dioxide

Table 2 provides hematologic, enzymatic, and injury-related variables. Dialysis patients exhibited higher hemoglobin [13.86 ± 3.85 vs. 12.42 ± 3.01 g/dL; Δ 1.44 (0.15 to 2.74), p = 0.029] and hematocrit [40.07 ± 10.93% vs. 35.40 ± 7.71%; Δ 4.68 (0.26 to 9.09), p = 0.039]. Lymphocyte count was lower [1.54 ± 0.71 vs. 1.86 ± 0.73 × 10³/µL; Δ − 0.32 (–0.62 to − 0.02), p = 0.040], while neutrophil count and NLR were significantly higher (both p < 0.001). Markers of muscle and hepatic injury—including creatine kinase [64,749 vs. 7,054 U/L], AST [880 vs. 167 U/L], ALT [385 vs. 96 U/L], and LDH [1,909 vs. 550 U/L]—were markedly elevated in the dialysis group (all p < 0.001). Rates of compartment syndrome (57.1% vs. 25.2%, p = 0.002), fasciotomy (35.7% vs. 8.9%, p = 0.001), and amputation (28.6% vs. 11.1%, p = 0.034) were also significantly higher.

Table 2.

Hematologic, enzymatic, and injury characteristics according to Dialysis requirement

Variable	No Dialysis (n = 135)	Dialysis (n = 28)	Δ Mean (95% CI) / Median [IQR]	p
Hemoglobin, g/dL	12.42 ± 3.01	13.86 ± 3.85	1.44 (0.15 to 2.74)	0.029
Hematocrit, %	35.40 ± 7.71	40.07 ± 10.93	4.68 (0.26 to 9.09)	0.039
Lymphocyte count, 10³/µL	1.86 ± 0.73	1.54 ± 0.71	–0.32 (–0.62 to − 0.02)	0.040
Neutrophil count, 10³/µL	11.21 [7.95–14.90]	15.51 [11.63–20.45]	–	0.001
Platelet count, 10³/µL	258.21 ± 86.84	231.93 ± 97.64	–	0.156
Mean platelet volume, fL	10.65 ± 1.00	10.82 ± 1.06	–	0.426
NL	6.20 [4.48–9.91]	11.08 [8.34–12.49]	–	< 0.001
PLR	140.72 [106.51–184.05]	162.34 [120.17–203.99]	–	0.242
Creatine kinase, U/L	7,054 [2,654–24,298]	64,749 [38,060–106,155]	–	< 0.001
AST, U/L	167 [68–385]	880 [477–1,284]	–	< 0.001
ALT, U/L	96.2 [43–180]	385 [267–582]	–	< 0.001
LDH, U/L	550 [397–844]	1,909 [1,189–2,836]	–	< 0.001
Trauma site count	1 [1–1]	1 [1–2]	–	0.003
Compartment syndrome, n (%)	34 (25.2%)	16 (57.1%)	–	0.002
Fasciotomy, n (%)	12 (8.9%)	10 (35.7%)	–	0.001
Amputation, n (%)	15 (11.1%)	8 (28.6%)	–	0.034

ALT, alanine aminotransferase; AST, aspartate aminotransferase; CK, creatine kinase; LDH, lactate dehydrogenase; NLR, neutrophil-to-lymphocyte ratio; PLR, platelet-to-lymphocyte ratio; SD, standard deviation; IQR, interquartile range

Table 3 compares diagnostic performance metrics. SAFE-QUAKE demonstrated a higher AUROC [0.894 (95% CI 0.831–0.957)] than the Dialysis Score [0.836 (95% CI 0.738–0.934)] (Fig. 1). Sensitivity at the Youden cutoff was higher for SAFE-QUAKE [82.1% vs. 69.2%], while specificity was higher for the Dialysis Score [90.6% vs. 84.4%]. The + LR was 5.28 for SAFE-QUAKE and 7.34 for the Dialysis Score; –LRs were 0.21 and 0.34, respectively. A formal DeLong test comparing AUCs showed a non-significant difference [Z = 1.415, p = 0.157], with an estimated AUC difference of 0.065 (95% CI − 0.025 to 0.154).

Table 3.

Comparative diagnostic performance of SAFE-QUAKE and Dialysis score for predicting Dialysis requirement

Metric	SAFE-QUAKE	Dialysis Score
AUROC (95% CI)	0.894 (0.831–0.957)	0.836 (0.738–0.934)
Youden Index Cutoff (95% CI)	≤ 4	≥ 2
Sensitivity at Youden (95% CI)	82.1% (67.9–96.7%)	69.2% (52.0–88.9%)
Specificity at Youden (95% CI)	84.4% (73.0–91.0%)	90.6% (66.0–96.3%)
+LR at Youden (95% CI)	5.28 (2.55–9.64)	7.34 (2.57–19.00)
–LR at Youden (95% CI)	0.21 (0.04–0.38)	0.34 (0.12–0.53)

AUROC – Area Under the Receiver Operating Characteristic Curve; CI – Confidence Interval; +LR – Positive Likelihood Ratio; –LR – Negative Likelihood Ratio; Youden Index – The threshold that maximizes (sensitivity + specificity – 1); SAFE-QUAKE – Seismic Assessment of Kidney Function to Rule Out Dialysis Requirement

Fig. 1.

Receiver operating characteristic curves of SAFE-QUAKE and dialysis scores for predicting dialysis requirement

The calibration performance of both models was also evaluated. While both showed acceptable agreement between predicted and observed dialysis probabilities, SAFE-QUAKE demonstrated slightly better calibration across probability ranges (Fig. 2).

Fig. 2.

Calibration plots of predictive models

Table 4 presents the category-free Net Reclassification Index (NRI), which favored the SAFE-QUAKE score [overall NRI 0.491 (95% CI 0.067–0.911)]. The entire reclassification benefit was driven by correct downward reclassification among non-dialysis patients [NRI⁻ = 0.491 (0.321–0.643)], while there was no net gain among dialysis-requiring patients [NRI⁺ = 0.000 (–0.404 to 0.410)]. Specifically, SAFE-QUAKE correctly reclassified 74.5% of controls downward and only 25.5% upward, yielding a more conservative risk estimate compared to the Dialysis Score (Fig. 3).

Table 4.

Category-free net reclassification improvement (NRI) comparing SAFE-QUAKE and Dialysis score

Metric	Estimate (95% CI)
Net Reclassification Index (NRI)	0.491 (0.067–0.911)
Positive NRI (NRI⁺)	0.000 (–0.404–0.410)
Negative NRI (NRI⁻)	0.491 (0.321–0.643)
↑ Reclassification in Cases	0.500 (0.298–0.705)
↓ Reclassification in Cases	0.500 (0.295–0.702)
↓ Reclassification in Controls	0.745 (0.660–0.821)
↑ Reclassification in Controls	0.255 (0.179–0.340)

↑: upward reclassification; ↓: downward reclassification

Fig. 3.

Category-based and category-free net reclassification improvement plots comparing safe-quake score with dialysis score interpretive notes: Left: Category-based NRI with a 0.5 risk threshold. Right: Category-free NRI without thresholds. Red dots represent cases (dialysis required); black dots represent controls. Points above the diagonal in the right panel indicate upward reclassification by the SAFE-QUAKE score compared to the Dialysis Score. Upward reclassification indicates patients who were correctly reclassified into a higher risk group by the second score, improving clinical sensitivity. Downward reclassification reflects patients reclassified into a lower risk category, which may suggest overestimation by the comparator model

Figure 4 displays prevalence-weighted PPV and NPV curves. At a hypothetical prevalence of 25%, SAFE-QUAKE yielded a PPV of 63.7% and NPV of 93.4%, while the Dialysis Score yielded a higher PPV of 71.0% but lower NPV of 89.8%. This trade-off remained consistent across prevalence values from 5 to 50%: SAFE-QUAKE showed superior NPV at all levels (e.g., 98.9% vs. 98.2% at 5% prevalence), while the Dialysis Score consistently demonstrated higher PPV (e.g., 28% vs. 22% at 5%, 88.0% vs. 84.0% at 50%).

Fig. 4.

Prevalence-weighted positive and negative predictive values of the safe-quake and dialysis scores. Interpretive note: Positive predictive value (PPV) reflects the likelihood that patients identified as high-risk truly required dialysis. Negative predictive value (NPV) reflects the probability that patients identified as low-risk did not require dialysis. Shifts in PPV/NPV illustrate the clinical impact of applying different scores on decision thresholds

Discussion

In our study, we examined the utility of two scoring systems—SAFE-QUAKE and the Dialysis Score—developed to predict dialysis requirement in patients presenting with earthquake-related injuries to the ED after the February 6 Kahramanmaraş earthquakes. We evaluated their performance in a center independent from those in which the scores were developed. We observed that SAFE-QUAKE serves as a safer triage tool, particularly for ruling out dialysis requirements, whereas the Dialysis Score provides a more accurate prediction of actual dialysis need.

The SAFE-QUAKE score was developed by Yılmaz et al. in one of the provinces affected by the earthquake [12]. The aim of that study was to identify early predictors of dialysis indication in patients presenting with earthquake-related injuries and to develop a quick and easy rule-out scoring system based on these factors. The SAFE-QUAKE score incorporated key predictors such as entrapment duration (< 45 h), pH level (> 7.31), creatinine level (< 2 mg/dL), LDH level (< 1600 U/L), and AST/ALT ratio (< 2.4). This approach aimed to help identify patients who do not require dialysis during the initial ED presentation using simple biochemical parameters and history, and to guide inter-facility transfers accordingly. In our study center, the SAFE-QUAKE score outperformed the Dialysis Score in ruling out dialysis need.

The Dialysis Score was developed by Comoglu et al. to predict which patients might require renal replacement therapy (RRT) during the initial ED assessment following crush injuries [13]. Unlike SAFE-QUAKE, the Dialysis Score was designed as a predictive tool rather than a rule-out score. It was based on the number of traumatized extremities, albumin levels, and CK levels. They argued that by assessing physical findings and biochemical markers, the dialysis requirement could be accurately predicted. Our study similarly observed that the Dialysis Score was more effective than SAFE-QUAKE in identifying patients who would actually require dialysis.

Following the 1999 Marmara Earthquake in Türkiye—magnitude 7.6, duration 45 s, with 17,480 deaths and 23,781 injuries—the national guideline “Hospital Management and Treatment Protocol in Mass Casualty Incidents” outlined dialysis indications: (1) BUN > 100 mg/dL or serum creatinine > 8 mg/dL; (2) serum potassium > 7 mmol/L; (3) blood pH < 7.1; (4) fluid overload; (5) clinical symptoms of uremia (e.g., uremic pericarditis, altered mental status, persistent nausea/vomiting); (6) prophylactic dialysis in crush syndrome even without classical indications [14]. These broad indications were primarily designed for the inpatient management of victims. In natural hazards like major earthquakes, 3–20% of mass casualties are due to crush injuries resulting from building collapse and entrapment [15, 16]. Although crush injuries are not limited to earthquakes, they are commonly encountered in the ED due to industrial, construction, or agricultural accidents involving entrapment.

The experience from the February 6 earthquakes has prompted emergency medicine researchers to re-evaluate triage approaches for patients with crush syndrome and potential acute kidney injury (AKI), particularly in resource-limited post-disaster settings. Both SAFE-QUAKE and the Dialysis Score, proposed by emergency medicine researchers to support dialysis decision-making in the ED, did not cover the full spectrum of dialysis indications listed in national or international guidelines [12, 13]. These guidelines seek to answer “When should dialysis be initiated?” while the scores aim to rapidly answer “Will this patient need dialysis?” [14, 17]. Disaster research emphasizes a triage-oriented approach throughout the entire patient journey—from field to discharge—by integrating medical and surgical care management [18–20]. Moreover, AKI management has long been acknowledged as a major determinant of mortality in patients presenting to the ED with crush injuries after disasters and MCIs [21]. This underscores the potential divergence between the ED perspectives and traditional treatment hierarchies. Therefore, in catastrophic disasters, from an ED perspective, early identification of dialysis requirement in patients with crush injuries could be a rational strategy. Both the SAFE-QUAKE and Dialysis Scores may represent preliminary tools for refining this strategy.

In disasters that lead to catastrophic destruction, effective management of resources—such as hospital beds, medical supplies, and diagnostic and treatment processes—is of critical importance, particularly in emergency departments. Earthquakes are among the most significant of these catastrophic events. In our study, we compared the SAFE-QUAKE and Dialysis Scores and observed that these two scoring systems can be used individually or in combination for the management of patients with crush injuries brought to the emergency department in mass casualty incidents (MCIs) and disasters where resource management is crucial. If the ED is part of a hospital capable of performing dialysis, patients with a SAFE-QUAKE score of 0 can be transferred to non-dialysis centers after completing their initial emergency care, allowing the dialysis-capable center to reserve its critical resources for patients who may require dialysis. For patients with a SAFE-QUAKE score of 1 or higher, the Dialysis Score can be used to estimate how many are likely to require dialysis, enabling strategic planning and resource allocation. This approach allows dialysis-capable centers to optimize their resources for the most critical patients during a disaster. Conversely, if the ED belongs to a hospital that cannot provide dialysis, patients with a SAFE-QUAKE score of 0 can be admitted locally, while those with a score of 1 or higher should be transferred to facilities with dialysis capabilities. In this way, using the SAFE-QUAKE and Dialysis Scores—either separately or in combination—can support the implementation of a strategic dialysis management algorithm in the aftermath of catastrophic earthquakes, such as the February 6 Kahramanmaraş earthquake, where the ability to perform dialysis plays a major role in reducing mortality among patients with crush injuries.

In such catastrophic events, the management capacity of affected regions is as important as the role of international emergency medical teams (EMTs), whose missions carry strategic significance [22]. EMTs serve as field-deployed EDs, and higher-level EMTs resemble fully structured health facilities. ED-focused disaster experiences can guide EMT development. The 2021 WHO classification for EMTs—outlined in the “Classification and Minimum Standards for Emergency Medical Teams”—categorizes them as Type 1 Mobile, Type 1 Fixed, Type 2, and Type 3 [23]. All EMT types possess the laboratory capacity needed to calculate SAFE-QUAKE and Dialysis Scores but only Type 3 teams can provide dialysis. Therefore, when planning interfacility transfers between EMTs, SAFE-QUAKE may guide referrals to Type 2 centers, while the Dialysis Score can be used to identify patients needing transfer to Type 3 centers. This can support more accurate bed management in resource-limited field hospitals and aligns with the “EMT 2030 Strategy” objective of optimizing resource use in disaster zones.

Looking ahead, the integration of artificial intelligence (AI) and machine learning algorithms could enhance the predictive performance of existing scoring systems such as SAFE-QUAKE and the Dialysis Score. By analyzing large-scale, multicenter datasets, AI models may capture complex, nonlinear relationships between clinical, laboratory, and injury-related variables, potentially offering real-time, adaptive decision support in disaster triage scenarios.

Limitations

This study was conducted in a province affected by the earthquake. Due to the nature of the disaster, patients with incomplete documentation could not be included. The study population consisted solely of patients who presented to the ED with crush injuries caused by the February 6 earthquakes in Türkiye, and it was assumed that dialysis decisions were made by clinicians in accordance with national and international guidelines. The analysis was limited to the comparison of only two scoring systems—SAFE-QUAKE and the Dialysis Score—without including other potential predictive models or artificial intelligence-based approaches.

Conclusion

Disasters and MCIs are situations in which prospective research is limited, the number of injured is high, and management algorithms are often shaped by experience. In this study, we compared the use of the Dialysis Score and the SAFE-QUAKE score and suggest that in future earthquakes and MCIs involving crush injuries: (1) If the continuation of treatment and management is to take place at a healthcare facility without dialysis capability, patient referral should be guided by the SAFE-QUAKE score in the ED. Conversely, if the continuation of care is to occur in a center with dialysis capability, patients with a high Dialysis Score should be identified and managed accordingly; (2) Considering varying patient volumes, the results suggest that SAFE-QUAKE may be particularly suitable in settings with low patient density, as it can be used safely to avoid unnecessary referrals. In contrast, in large hospitals with high patient load, the Dialysis Score may be more effective in identifying those who truly require dialysis; (3) SAFE-QUAKE has demonstrated high accuracy in correctly identifying patients who do not require dialysis, thereby enabling safe management of a substantial proportion of patients presenting after a disaster. This may help reduce unnecessary referrals and ease the burden on the healthcare system. Especially in resource-limited disaster settings, such a triage or decision-making tool may enhance the efficiency of patient management.

Author contributions

SY, SA, ÇSB, AY, NES and ACT conceived the study, designed the trial. RC, MO, ACT, and AS supervised the conduct of the trial and data collection. ÇSB, and AY undertook recruitment of participating centers and patients and managed the data, including quality control. SY, SA, and ACT, NES provided statistical advice on study design and analyzed the data; SY, SA, ÇSB, and AY chaired the data oversight committee. SY, SA, ÇSB, AY, NES and ACT drafted the manuscript, and all authors contributed substantially to its revision. SY, SA, ÇSB, AY, NES and ACT takes responsibility for the paper as a whole.

Funding

The authors received no financial support for the conduct, authorship, or publication of this research.

Data availability

The entire deidentified dataset, data dictionary and analytic code for this investigation are available upon request, from the date of article publication by contacting Sarper YILMAZ, Assoc. Prof., at [sarperyilmaz08@gmail.com](mailto: sarperyilmaz08@gmail.com).

Declarations

Ethics approval and consent to participate

Ethical approval for this study was obtained from the Mersin University Clinical Research Ethics Committee prior to study initiation (Approval Date: 30.04.2025, Approval No: 452). All procedures were conducted in accordance with the principles of the Declaration of Helsinki.

Consent for publication

Due to the retrospective nature of the study and the use of anonymized patient data, the requirement for informed consent was waived by the ethics committee.

Competing interests

The authors declare no competing interests.