a b s t r a c t

Objective: To establish and validate a predictive formula for calculating the possibility of developing delayed neu- rological sequelae (DNS) after Acute carbon monoxide poisoning to facilitate better decision-making about treatment strategies.

Methods: This study retrospectively enrolled 605 consecutive patients who had been newly diagnosed with CO poisoning from the Central Hospital of Enshi Prefecture between January 1, 2015 and December 31, 2020. The co- hort was randomly divided into two subgroups: the development cohort (n = 104) and validation cohort (n = 44). Univariate analysis and backward elimination of multivariate logistic regression were used to identify pre- dictive factors, and a predictive formula was established. The performance was assessed using the area under the curve (AUC), the mean AUC of five-fold cross-validation, and calibration plots.

Results: The formula included four commonly available predictors: initial GCS score, duration of exposure, CK, and abnormal findings on MRI. We next created a formula to calculate the risk score for developing DNS: Risk score =

-4.54 + 3.35 * (Abnormal findings on MRI = yes) - 0.51 * (Initial GCS score) + 0.65 * (Duration of expo-

sure) + 0.01 * (CK). Then, the probability of developing DNS could be calculated: Probability of DNS = 1/ (1 + e ^{Risk score}). The model revealed good discrimination with AUC, and mean AUC of fivefold cross-validation in two cohort, and the calibration plots showed good calibration.

Conclusions: This study established a prediction predictive formula for predicting developing of DNS, which could facilitate better decision-making about treatment strategies.

(C) 2021 Published by Elsevier Inc.

1. Introduction

Carbon monoxide poisoning, which causes dysfunction of mul- tiple organs, particularly within the cardiovascular and central nervous systems, is a significant cause of disability and death [1]. CO binds tightly to hemoglobin and forms carboxyhemoglobin (COHb) in the blood, which may cause tissue and cell hypoxia [2]. Because the brain and heart are more sensitive to hypoxia related to high oxygen demands, these organs are most vulnerable to CO poisoning. Symptoms of CO poi- soning are nonspecific, but can include headaches, myalgia, chest pain, dizziness, Neuropsychological impairments, and even death [3].

Delayed neurological sequelae (DNS) is a serious complication that occurs between several days and six weeks after successful initial

Abbreviations: CO, Carbon monoxide; COHb, Carboxyhemoglobin; DNS, Delayed neurological sequelae; HBO, Hyperbaric oxygen; GCS, Glasgow Coma Scale; NSE, Neuron-specific enolase; CK, Creatine kinase; LDH, Lactate dehydrogenase; WBC, White blood cell; MRI, Magneticresonanceimaging; ROC, Receiver operator characteristic curves; AUC, Area under the curve; DWI, diffusion-weighted imaging.

* Corresponding author.

E-mail address: [email protected] (J. Li).

¹ These authors contributed equally to the manuscript.

resuscitation from CO acute intoxication. Its incidence is about 15%-40% [4]. However, the pathogenesis of DNS after Acute CO poisoning remains unknown. Perhaps inflammatory and mitochondrial oxidative phosphor- ylation produce ischemic and Anoxic brain injury, which leads to NDS [5]. Identifying patients with acute CO poisoning who are likely to de- velop DNS is a major challenge. Numerous predictors relevant to DNS, in- cluding loss of consciousness, longer-duration CO exposure, serum lactate, cardiac enzyme, leukocyte, and Troponin I levels, and abnormal imaging findings, have been identified in previous studies [1,4-7]. Al- though these predictors were identified, however, there are no Predictive tools for integrating multiple variable and quantitative analyses. The pur- pose of this study was to identify factors and generate predictive models for the development of DNS after acute CO poisoning. We hope this will

help facilitate better decision-making about treatment strategies.

1. Materials and methods
   1. Study population

Our study involved patients with CO poisoning from a cohort from the Central Hospital of Enshi Prefecture who were treated between

https://doi.org/10.1016/j.ajem.2021.10.056 0735-6757/(C) 2021 Published by Elsevier Inc.

January 1, 2015 and December 31, 2020. We retrospectively collected data on 605 patients who had been newly diagnosed with CO poisoning. The inclusion criteria were as follows: (i) patients newly diagnosed with acute CO poisoning, (ii) undergone regular follow-up every 1-3 months in outpatient clinics or by telephone. Exclusion criteria were:

(i) patients aged <18 years, (ii) patients who did not receive Hyperbaric oxygen therapy, (iii) patients who had cardiac arrest, (iv) pa- tients who were discharged against medical advice, (v) patients who had neurological symptoms before CO intoxication and (vi) patients who failed to follow-up. A final total of 148 patients were included in the study (Appendix 1). This study was approved by the Ethics Commit- tee of the Central Hospital of the Enshi Prefecture. All participants pro- vided written informed consent.

• 1. Data collection

We retrospectively examined all patients’ medical histories, includ- ing sex, age, Glasgow Coma Scale score on arrival, duration of CO exposure, COHb value, arterial pH, troponin I, neuron-specific eno- lase (NSE), creatine kinase (CK), Lactate dehydrogenase , lactate, the number of HBO sessions, white blood cell count, hemoglobin, and MRI findings.

Blood samples were taken from the patients within 24 h of the

acute CO poisoning onset. Routine blood tests (for WBC count and he- moglobin) levels were carried out using an automatic blood cell ana- lyzer (Sysmex, Kobe, Japan). Arterial blood gas levels, including COHb, arterial pH, and serum lactate, were measured using a pulse co-oximeter (Masimo, Irvine, USA). CK was measured using the color- imetric method with a creatine kinase activity assay kit (Abcam, Cam- bridge, USA), LDH was measured using an AU 2700 automatic biochemistry analyzer (Olympus, Tokyo, Japan). cTnI was measured using a chemiluminescent immunoassay on an Access 2 analyzer (Beckman Coulter, California, USA). NSE was measured with a solid- phase immunoassay with double monoclonal antibodies, using an Elecsys 2010 (Roche Diagnostics GmbH, Basel, Switzerland). Creatinine was measured by the Jaffe reaction that is kinetic colorimetric assay (Roche Diagnostics, Basel, Switzerland). MRI scans were performed on a 1.5 T unit (Siemens, Erlangen, Germany) scanner; The MRI proto-

tests. Backward elimination multivariable logistic regression was used to select predictors for the model. We performed a stepwise selection procedure using backward selection procedures: Backward elimination started with a sequence of tests to remove or keep variables in the model (based on p < 0.05 for variable exclusion). Odds ratios and the 95% confidence intervals (95% CI) were calculated as well. A P < 0.05 was considered statistically significant for all tests. Finally, we created a formula to calculate risk scores and calculate the probability of devel- oping NDS.

Discrimination and calibration were used to evaluate the predictive accuracy of the established prediction model. The discriminatory ability of the model was evaluated using receiver operator characteristic curves (ROC), sensitivity, specificity, negative likelihood ratios, and pos- itive likelihood ratios. Calibration was assessed using calibration plots, which measured the relationship between predicted probabilities and observed proportions. We also assessed generalizability using the mean the area under the curve (AUC) with stratified fivefold cross- validation.

Statistical analyses were conducted using statistical software (R statistical software v3.5.2) with the ‘pROC’, ‘rms’, and ‘rpart’ pack- ages.

1. Results

A total of 605 patients who had been newly diagnosed with CO poisoning were retrospectively enrolled in the study. After ruling out patients who did not meet the inclusion criteria, we included 148 patients, of whom 104 were randomly assigned to the develop- ment cohort and 44 were randomly assigned to the validation cohort. Detailed clinical and demographic characteristics for both groups are shown in Table 1.

Nine candidate predictors were identified in the development co- hort following univariate analysis (Table 2). After backward elimination

Table 1

Clinical and demographic characteristics of the development cohort and validation cohort.

col consisted of diffusion-weighted imaging (DWI) scans with fluid- attenuated inversion recovery imaging (FLAIR). All MRI images were analyzed by at least two experienced radiologists, who were blinded to the patients’ clinical information. Discrepancies in judging the oc- currence of abnormalities were settled by joint discussion between the two radiologists.

Variable Development cohort

(n = 104)

Age, years Initial GCS score	55.5(47-67) 12(9-13)	64(50.5-72) 12(10-14)	63(49-71) 12(10-14)
Duration of exposure, h	7.5(6.35-10.35)	7.1(6-8.6)	7.2(6-8.6)
NSE, ng/mL	5.55(3.3-7.7)	5.6(3.25-7.4)	5.6(3.3-7.45)
Arterial pH	7.39(7.37-7.4)	7.39(7.37-7.41)	7.39(7.37-7.41)
Lactate, mmol/L	1.3(0.86-1.59)	1.3(0.9-1.7)	1.3(0.9-1.6)
Creatinine, umol/L	66.85(53.7-75.5)	62.25(49.85-81)	64.7(51.1-80.05)
CK, U/L	99(76-177.5)	105.5(70-168)	103.5(71-170.5)
LDH, U/L	166.75 +- 53.48	172.13 +- 57.59	170.53 +- 56.27
White blood cell, 109/L	7.05(5.37-8.72)	6.3(5.06-7.95)	6.31(5.17-8.14)
Hemoglobin, g/L	126(115-145)	127(118.5-140.5)	127(116.5-142)
Number of HBO sessions Sex	7(7-10)	7(7-11)	7(7-11)
Female	20(45.45)	61(58.65)	81(54.73)
Male	24(54.55)	43(41.35)	67(45.27)

Validation cohort (n = 44)

Total

(n = 148)

2.3. Definitions

DNS was defined as any neurological sign that developed within 3 months after CO intoxication. Those signs included motor deficits, cog- nitive decline, dysphagia, seizures, neuropsychological disease, extrapy- ramidal symptoms and incontinence. The normal range was defined as

7.35-7.45 for arterial pH, 0.5-2 mmol/L for lactate, 35-110umol/L for creatinine, 24-195 U/L for CK, 50-250 U/L for LDH, (4.0-10.0) x 10^9/L for WBC, 120-160 g/L for hemoglobin, and 0-12.5 ug/L for NES. Abnor- malities on MRI scans included globus pallidus lesions, diffuse lesions, and focal lesions (territorial lesions, patchy lesions, punctate lesions) [7].

	Negative	43(97.73)	102(98.08)	145(97.97)
	Positive	1(2.27)	2(1.92)	3(2.03)
2.4. Statistical analysis	Troponin I
	Negative	44(100)	100(96.15)	144(97.3)

COHb >=25%

No 31(70.45) 90(86.54) 121(81.76)

Yes 13(29.55) 14(13.46) 27(18.24)

Myoglobin

Normally distributed continuous variables were expressed as mean +- standard deviation, and assessed using t-tests between the two groups (non-DNS and DNS). Non normally-distributed continuous variables were presented as medians and quartiles, and were assessed with Wilcoxon tests. Categorical data were expressed as frequencies and percentages, and were assessed using ?2 tests or Fisher’s exact

Positive 4(3.85) 4(2.7)

Abnormal findings on MRI

No 34(77.27) 82(78.85) 116(78.38)

Yes 10(22.73) 22(21.15) 32(21.62)

Abbreviations: GCS, Glasgow Coma Scale; NSE, Neuron-specifc enolase; CK, Creatine ki- nase; LDH, lactate dehydrogenase; HBO, hyperbaric oxygen; COHb, carboxyhemoglobin; MRI: Magnetic resonance imaging.

Table 2

Univariable logistic regression to select candidate variables in the development cohort.

Variable	Non-DNS (n = 82)	NNS (n = 22)	P
Age, years	65(48-74)	56(55-66)	0.358
Initial GCS score	13(12-14)	8.5(8-10)	0.000
Duration of exposure, h	6.9(5.6-7.6)	9.9(8.6-15.2)	0.000
NSE, ng/mL	5.1(2.9-7.2)	6.35(4.3-9.1)	0.039
Arterial pH	7.4(7.38-7.41)	7.38(7.36-7.42)	0.309
Lactate, mmol/L	1.4(0.9-1.7)	1.2(0.95-1.6)	0.526
Creatinine, umol/L	59.2(47.9-76)	77.5(62.2-91)	0.010
CK, U/L	95.5(68-125)	611(206-1120)	0.000
LDH, U/L?	178.78 +- 53.73	147.32 +- 65.68	0.022
White blood cell, 109/L	6.27(5-7.7)	6.47(5.6-8.1)	0.450
Hemoglobin, g/L	126(115-138)	132.5(124-142)	0.033
Number of HBO sessions	7(7-11)	8(7-11)	0.562
Sex			0.157
Female	51(62.2)	10(45.45)
Male COHb >=25%a	31(37.8)	12(54.55)	0.000
No	79(96.34)	11(50)
Yes Myoglobina	3(3.66)	11(50)	0.380
Negative	81(98.78)	21(95.45)
Positive Troponin Ia	1(1.22)	1(4.55)	1.000
Negative	79(96.34)	21(95.45)
Positive Abnormal findings on MRIa	3(3.66)	1(4.55)	0.000
No	76(92.68)	6(27.27)
Yes	6(7.32)	16(72.73)

P-value < 0.05 was considered statistically significant. Abbreviations: P, P value; GCS, Glasgow Coma Scale; NSE, Neuron-specifc enolase; CK, Cre- atine kinase; LDH, lactate dehydrogenase; HBO, hyperbaric oxygen; COHb, carboxyhemoglobin; MRI: Magnetic resonance imaging.

* Data are means +- standard deviation (SD).

^a Fisher’s exact probability method was used.

of variables in the multivariable logistic regressive analysis, four predic- tors remained in the final model (Table 3), including initial GCS score, duration of exposure, CK, and abnormal findings on MRI. We next cre- ated a formula to calculate the risk score for developing DNS: Risk score = -4.54 + 3.35 * (Abnormal findings on MRI = yes) - 0.51 * (Ini- tial GCS score) + 0.65 * (Duration of exposure) + 0.01 * (CK). Then, the probability of developing DNS could be calculated: Probability of DNS = 1/(1 + e ^{Risk score}).

The model showed good discrimination. In the development cohort, the AUC, sensitivity, specificity, negative likelihood ratio and positive likelihood ratio of this predictive formula were 0.99, 0.97, 0.90, 4.24, 32.91, and 9.81, respectively. In the validation cohort, the AUC, sensitivity, specificity, negative likelihood ratio and positive likelihood ratio of this model were 0.98, 0.94, 0.90, 4.24, 16.45, and 9.50, respectively. (Fig. 1).

Calibration plots indicated a good fit of the predicted probabilities and observed proportions (Fig. 2). The mean AUC derived from

Table 3

Multivariable logistic regression to select the final predictive factors in the development

cohort.
Variable	Regression coefficient	Standard error	Wald x2	P	OR (95%CI)
Intercept	-4.542	3.859	1.385	0.239
Initial GCS score	-0.510	0.255	3.996	0.046	0.60(0.36-0.99)
Duration of exposure, h	0.651	0.319	4.159	0.041	1.92(1.03-3.59)
CK, U/L	0.008	0.005	2.317	0.049	1.01(1-1.02)
Abnormal findings on MRI Yes	3.347	1.354	6.110	0.013	28.42(2-403.70)
No

P-value < 0.05 was considered statistically significant.

Abbreviations: P, P value; GCS, Glasgow Coma Scale; CK, Creatine kinase; MRI: Magnetic resonance imaging.

stratified fivefold cross-validation were 0.99 in the development cohort and 0.98 in the validation cohort, indicating good generalizability.

1. Discussion

In our study, we developed and validated a predictive formula for the probability of the development of DNS in patients who were newly diagnosed with acute CO poisoning. The four predictors were ini- tial GCS score, duration of exposure, CK, and abnormal findings on MRI, all of which have previously been reported as predictors of DNS. The probability of developing DNS could then be calculated precisely based on the formula we established.

A lot of predictors based on clinical data and statistical analyses have been identified in previous studies. In a multicenter study, Zhang et al. demonstrated that a longer duration of CO, as well as lower GCS scores evaluated on arrival, were independent predictors of DNS. This study also showed that patients with CO exposure >4.8 h were significantly predisposed to DNS [8]. Weaver et al. also believed that duration of CO exposure was an important factor related to the development of DNS, and demonstrated that exposure intervals greater than or equal to 24 h led to increased chances for developing DNS [9]. Another study, conducted by Pepe et al., demonstrated that a duration of CO exposure

>6 h might have increased the risk of DNS [10]. According to this study, duration of CO exposure was a vital factor for the development of DNS. However, the exact length of CO exposure that leads to DNS is controversial, because different cut-off values have been found in differ- ent studies that originate from different research centers. Many pro- spective studies have found that a GCS score < 9 was a significant predictor of DNS development. [10,11]. GCS scores objectively reflect the patient’s state of consciousness, and a lower GCS score was a poor predictive factor. Although GCS scores at the time of patient presenta- tion to the emergency room are widely-used predictors of DNS, the cut-off values vary in different studies.

Previous studies also found that some serum markers, including CK, CK-MB, LDH, CRP, cardiac troponin I, COHb, and S100B protein levels, as well as leukocytosis, could be used to predict the occurrence of DNS [1,6,12]. In our study, we demonstrated that patients’ CK levels were an independent predictor of DNS, which is consistent with other studies [6,11]. CK reflects intracellular enzymes mainly located in skeletal mus- cle, myocardium, and the brain, and high CK levels can be related to CO- induced ischemia, inflammatory reactions, and long-Standing postures due to loss of consciousness. Thus, high CK levels suggest that CO poi- soning has caused damage in multiple organs, and can be used as pre- dictors for the development of DNS. Lee et al. [13] examined CK levels in these patients, and found that the normal reference of concentration ranged from 160 U/L to 200 U/L (and used this as the cut-off value).

Previous studies have demonstrated that MRI-DWI scans that showed acute ischemic brain lesions were significantly associated with the development of DNS; Abnormal changes such as white matter hyperintensities, hippocampal atrophy, globus pallidus lesions, puta- men lesions, thalamus lesions, caudate nucleus lesions and certain ab- normal white matter signals, were early predictors for the risk of developing DNS, and were helpful in treating patients with CO poison- ing [14,15]. Those lesions may destroy the cortex and the neurofibrillary network complex that connects the cortex, resulting in multiple cogni- tive impairments as well as balance and gait disorders.

CO poisoning is a potentially lethal condition, and nursing care is very important, especially in patients with children, adults with cardiac disease, pregnant women, patients with increased oxygen demand or decreased oxygen-carrying capacity, and/or patients with chronic respi- ratory insufficiency. Nurses should also monitor vital signs and carry out a high standard of basic nursing care, including helping patients with left limb Movement disorder avoid falling, decreasing the risk of aspira- tion for patients with impaired swallowing function, and employing turnovers to avoid the formation of bedsores in bedridden patients

Fig. 1. ROC of the predictive model in development cohort and validation cohort.

(A) ROC of development cohort; (B) ROC of validation cohort.

Abbreviations: ROC, Receiver operator characteristic curves; AUC, area under the curve.

Fig. 2. Calibration plots of the predictive model in development cohort and validation cohort.

(A) Calibration plot of development cohort; (B) Calibration plot of validation cohort. A 45 diagonal line indicates perfect calibration.

[16]. Patient education and risk assessments must also be part of the nurses’ intervention repertoire.

Our study has several strengths. First, we identified the predictors of DNS, and established and validated a comprehensive formula which can be used to calculate the probability of the occurrence of DNS after CO poi- soning. Second, this formula integrating multiple variables and con- ducting quantitative analysis, is novel, user-friendly, data-based and reliable. Third, we used multiple methods to determine the validity of the model.

Our study also had some limitations. First, our study was a single- center design based on retrospective data, so the results may not be generalizable to other centers, and a large-scale double-blinded ran- domized study is needed. Second, the predictive model is not suitable for the patients who did not undergo HBO therapy, patients who did not complete follow-up, or patients who were aged <18 years. Third, the CO exposure concentrations were different in patients, and different patients received different durations of oxygen therapy, which may have influenced the prognosis. Finally, differences in prehospital man- agement were not considered, although all patients underwent the same treatment protocol since being admitted to our hospital.

1. Conclusion

In summary, this study identified risk factors for the development of DNS after CO poisoning, and established and validated a predictive for- mula to calculate the probability for developing DNS.

Funding

This work was supported by the Science and Technology Bureau of Enshi Prefecture (Grant number JCY2019000008).

Authors’ contributions

YSJ, LHC and PQF were joint first authors; LJL was correspondence author; YSJ and LHC designed the study; YSJ, LHC, PQF and LJL reviewed the literature; YSJ and PQF collected the data; YSJ and PQF performed the follow-up activity; YSJ performed the statistical analysis; YSJ and LHC wrote the manuscript; LQH and LJL revised the manuscript. All au- thors have read and approved the final manuscript.

Declaration of Competing Interest

The authors declare that they have no conflict of interest.

Acknowledgment

We would also like to thank our staff, who assisted in the data collec- tion and analysis.

Appendix 1. Flow chart

Abbreviations: CO, Carbon monoxide; DNS, Delayed neurological se- quelae.

References

1. Weaver LK. Clinical practice. Carbon monoxide poisoning. N Engl J Med. 2009;360 (12):1217-25. https://doi.org/10.1056/NEJMcp0808891.
2. Hampson NB, Hauff NM. Risk factors for short-term mortality from carbon monox- ide poisoning treated with hyperbaric oxygen. Crit Care Med. 2008;36(9):2523-7. https://doi.org/10.1097/CCM.0b013e31818419d8.
3. Hampson NB, Piantadosi CA, Thom SR, Weaver LK, et al. Practice recommendations in the diagnosis, management, and prevention of carbon monoxide poisoning. Am J Respir Crit Care Med. 2012;186(11):1095-101. https://doi.org/10.1164/rccm. 201207-1284CI.
4. Rose JA-O, Wang LA-O, Xu Q, McTiernan CF, Shiva S, Tejero J, et al. Carbon monoxide poisoning: pathogenesis, management, and future directions of therapy. Am J Respir Crit Care Med. 2017;195(5):596-606. https://doi.org/10.1164/rccm.201606-1275CI.
5. Thom SR, Bhopale VM, Fisher D. Hyperbaric oxygen reduces delayed immune- mediated neuropathology in experimental carbon monoxide toxicity. Toxicol Appl Pharmacol. 2006;213(2):152-9. https://doi.org/10.1016/j.taap.2005.10.006.
6. Kim H, Choi S, Park E, Yoon E, Min Y, Lampotang S. Serum markers and development of delayed neuropsychological sequelae after acute carbon monoxide poisoning: anion gap, lactate, osmolarity, S100B protein, and interleukin-6. Clin Exp Emerg Med. 2018;5(3):185-91. https://doi.org/10.15441/ceem.17.217.
7. Jeon SB, Sohn CH, Seo DW, Oh BJ, Lim KS, Kang DW, et al. Acute brain lesions on magnetic resonance imaging and delayed neurological sequelae in carbon monoxide poisoning. JAMA Neurol. 2018;75(4):436-43. https://doi.org/10.15441/ceem.17. 217.
8. Zhang Y, Lu Q, Jia J, Xiang D, Xi Y. Multicenter retrospective analysis of the risk fac- tors for delayed neurological sequelae after acute carbon monoxide poisoning. Am J Emerg Med. 2020;10(689):1-11. https://doi.org/10.1016/j.ajem.2020.06.090.
9. Weaver LK, Valentine KJ, Hopkins RO. Carbon monoxide poisoning: risk factors for cognitive sequelae and the role of hyperbaric oxygen. Am J Respir Crit Care Med. 2007;176(5):491-7. https://doi.org/10.1164/rccm.200701-026OC.
10. Pepe G, Castelli M, Nazerian P, Vanni S, Del Panta M, Gambassi F, et al. Delayed neu- ropsychological sequelae after carbon monoxide poisoning: predictive risk factors in the emergency department. A retrospective study. Scand J Trauma Resusc Emerg Med. 2011;19:16. https://doi.org/10.1186/1757-7241-19-16.
11. Han S, Choi S, Nah S, Lee SU, Cho YS, Kim GW, et al. Cox regression model of prog- nostic factors for Delayed neuropsychiatric sequelae in patients with acute carbon monoxide poisoning: a prospective observational study. NeuroToxicology. 2021; 82:63-8. https://doi.org/10.1016/j.neuro.2020.11.006.
12. Runming S, Wenyuan C, Zhangge J, Wenwen B, Lixia W, Qian W, et al. Predictive values of serum biochemical markers and apparent diffusion coefficient on Delayed encephalopathy after acute carbon monoxide poisoning. Turk Neurosurg. 2020;5: 1-6. https://doi.org/10.5137/1019-5149.JTN.30830-20.2.
13. Lee H, Kang H, Ko BS, Oh J, Lim TH, Cho Y. Initial creatine kinase level as predictor for delayed neuropsychiatric sequelae associated with acute carbon monoxide poison- ing. Am J Emerg Med. 2021;43:195-9. https://doi.org/10.1016/j.ajem.2020.02.054.
14. Nah S, Choi S, Kim HB, Lee J, Lee SU, Lee YH, et al. Cerebral white matter lesions on diffusion-weighted images and delayed neurological sequelae after carbon monox- ide poisoning: a prospective observational study. Diagnostics (Basel). 2020;10(9): 1-9. https://doi.org/10.3390/diagnostics10090698.
15. Jeon SB, Sohn CH, Seo DW, Oh BJ, Lim KS, Kang DW, et al. Acute brain lesions on magnetic resonance imaging and delayed neurological sequelae in carbon monoxide poisoning. JAMA Neurol. 2018;75(4):436-43. https://doi.org/10.1001/jamaneurol. 2017.4618.
16. Ackley BJ, Ladwig GB. Nursing diagnosis handbook: A guide to planning care. 8th ed.. St Louis, MO: Mosby Elsevier; 2008; 148-55.