clinical decisionmaking based on venous versus capillary blood gas values in the well-perfused child

5 8 A N N A L S O F E M E R G E N C Y M E D I C I N E 3 4 : 1 J U L Y 1 9 9 9

David McGillivray, MD, FRCP(C) Francine M Ducharme, MD,

FRCP(C) Yves Charron, DEC, RTII

Claire Mattimoe, MD, FRCP(C)Steve Treherne, MD, FRCP(C)

From the Departments of Pediatrics,*

Epidemiology,‡ Biostatistics,§ andBiochemistry,II Montreal Children’sHospital, McGill University,Montreal, Quebec, Canada.

Received for publication April 22, 1998. Revision received January 4, 1999. Accepted for publication February 8, 1999.

Presented at the AmbulatoryPediatric Association Meeting,Seattle, WA, United States, May 1994, and the CanadianPediatric Society Meeting, Emergency Medicine Section, St John’s, Newfoundland, Canada,July 1994.

Address for reprints: Dr DavidMcGillivray, Montreal Children’sHospital, 2300 Tupper Street, A-103,Montreal, Quebec, Canada H3H 1P3;E-mail [email protected].

Copyright © 1999 by the AmericanCollege of Emergency Physicians.

0196-0644/99/$8.00 + 047/1/97855

P E D I A T R I C S / O R I G I N A L C O N T R I B U T I O N

Clinical Decisionmaking Based on Venous

Versus Capillary Blood Gas Values in the Well-

Perfused Child

Study objective: In children aged 1 month to 18 years, wesought to examine the correlation between venous and arterial-ized capillary blood gas values, and to determine whether thesource of blood sample influenced the interpretation of theacid-base status and clinical management.

Methods: In a cross-sectional study, venous and capillaryblood gas values were simultaneously obtained in acutely illwell-perfused patients treated in a pediatric emergency depart-ment. Intraclass correlation coefficients for capillary andvenous measured gas values were calculated. Crude agreementand intraobserver concordance were calculated for responsesof 2 intensivists to the interpretation and clinical managementquestions, based on capillary and venous gas results.

Results: Intraclass correlation coefficients for 78 capillary andvenous paired measured gas values were .92 (pH), .80 (PCO2),and .67 (PO2). The α of concordance values between capillaryand venous blood gas values, with 95% confidence intervals(CIs) were as follows, respectively, for physician A and B: inter-pretation, .61 (.47 to .73) and .48 (.41 to .55); need for bicarbon-ate, .85 (.73 to .97) and .80 (.72 to .88); and need for intubation.73 (.64 to .82), and .83 (.75 to .91).

Conclusion: In the well-perfused patient, we believe thatvenous samples are an acceptable alternative to capillary bloodsamples for determination of blood gas values and for makingclinical management decisions.

[McGillivray D, Ducharme FM, Charron Y, Mattimoe C, TreherneS: Clinical decisionmaking based on venous versus capillaryblood gas values in the well-perfused child. Ann Emerg MedJuly 1999;34:58-63.]

D E C I S I O N M A K I N G O N V E N O U S V E R S U S C A P I L L A R YB L O O D G A S V A L U E S

McGillivray et al

J U L Y 1 9 9 9 3 4 : 1 A N N A L S O F E M E R G E N C Y M E D I C I N E 5 9

capillary, and venous samples.6-15 However, there are nopublished studies comparing a large number of simulta-neously obtained capillary gas samples with venous gassamples throughout the pediatric age range to examinehow acid-base values correlate or whether the samplesource influences the patient’s clinical management.

A large group of children who are mildly to moderatelyill can benefit from acid-base status determination forassisting in diagnosis, management, and progression ofillnesses such as diabetic ketoacidosis, intoxications,altered level of consciousness, bronchiolitis, croup,asthma, metabolic disorders, renal disorders, and mod-erate dehydration. Given the simplicity and practicalityof venous procurement, we wished to study the value ofvenous gas samples in well-perfused patients aged 1month to 18 years to assess acid-base status and guideclinical management. Our specific objectives were asfollows: (1) to examine the correlation between bloodgas values simultaneously obtained from capillary andvenous samples, and (2) to determine whether the sam-pling source (capillary versus venous) would alter thepatient’s clinical management.

M A T E R I A L S A N D M E T H O D S

We recruited a convenience sample of acutely ill patientsrequiring acid-base determination who presented during

I N T R O D U C T I O N

Arterial gas sampling represents the gold standard fordetermination of a patient’s acid-base status. Risks asso-ciated with arterial puncture include pain, damagedartery, and loss of function of the extremity, hemorrhage,aneurysmal formation, or thrombosis of the artery.1-5

These risks increase with repeated arterial punctures,in children with small arteries, with insertion of acatheter, and when performed by individuals with lim-ited experience. Capillary gas sampling is a less inva-sive way of evaluating acid-base status in the well-per-fused patient. As with arterial gas sampling, capillarytechnique requires a separate puncture. Most patientsrequiring assessment of the acid-base status will have avenipuncture for other laboratory tests or venous access.If a blood gas value determined from venous samplescould be used to guide patient management with thesame accuracy as capillary or arterial samples, a singlevenipuncture could allow the sampling for blood gasvalues, laboratory tests, and intravenous insertion.

For many years clinicians and researchers have beenlooking for alternatives to arterial gas sampling in bothchildren and adults. Studies comparing arterial and capil-lary blood samples in diabetic ketoacidosis, arterial andvenous blood gas values in diabetic ketoacidosis, andarterial and capillary blood gas values in stable pediatricICU patients have shown good correlation among arterial,

Figure 1.Age distribution of partici-pants.


McGillivray et al


sures.17,18 Crude agreement and the α of concordancewith 95% confidence intervals (CIs) were determined toassess intraobserver agreement for the interpretive andmanagement questions. The α of concordance was cho-sen as measure of agreement, because, in contrast withpure agreement, it corrects for chance while being lessconservative than κ.19 Concordance values of agreementgreater than .75 may be taken to represent excellentagreement, values below .40 poor agreement, and valuesbetween .40 and .75 may be taken to represent fair togood agreement beyond chance.20

R E S U L T S

Seventy-eight patients aged 1 month to 18 years wereenrolled in the study (Figure 1). The most common spe-cific diagnoses included known or newly diagnosed dia-betes mellitus (N=9), moderate respiratory distress(asthma, bronchiolitis, or pneumonia; N=22), vomitingwith or without diarrhea (N=19), seizures (N=7), anddrug overdose (N=12). Seventy-five percent of the capil-lary blood gas samples were abnormal according to thereference values of pH, PCO2, and HCO3 proposed by thePediatric Advanced Life Support Guidelines (Table 1).21

One of the 78 patients was intubated and admitted to thepediatric ICU and 2 patients received NaHCO3 therapy.

With the exception of PO2, venous and capillary gasvalues showed a strong linear relationship, with intraclasscorrelation coefficients greater than or equal to .80(Figure 2).

Crude agreement for all the responses comparing cap-illary-venous samples by both physicians were greaterthan or equal to .59 for gas interpretation, greater than orequal to .91 for decisions regarding the need for bicar-bonate or intubation, and greater than or equal to .67 fordecisions regarding the need for change in ventilation ifthe patients were ventilated. For the 24 capillary gases

a 12 month period to the medical emergency departmentof a tertiary care pediatric teaching hospital. The designincluded a cross-sectional sampling of capillary andvenous gases that were subsequently submitted, with abrief written history, to 2 intensivists for interpretationand management decisions. Although the 2 intensivistsworked in the same hospital, they trained in differentcountries. The protocol was reviewed and approved bythe institutional review board. No informed consent wasobtained because no experimental treatment was under-taken and the procurement of specimens was part of thenecessary patient care.

Eligibility criteria included the following: age 1 monthto 18 years; normal vascular perfusion as assessed byheart rate, blood pressure, capillary refill, and level ofconsciousness; need for acid-base determination; needfor a venipuncture either for laboratory tests or intra-venous access. The last 3 criteria were determined by thetreating physician and independently confirmed by theprimary investigator (DM). Venous and capillary bloodsamples for determination of blood gas values were takenwithin 5 minutes of each other. Venous samplinginvolved drawing the blood from a free-flowing venousaccess without a tourniquet in place. Capillary samplingwas done by fingerprick technique in a well-perfuseddigit.16 Two of 4 pediatric intensivists (CM and ST) fromour pediatric ICU agreed to interpret blood gas values andmake management decisions based on the results. A briefwritten clinical history and physical examination, writtenby the primary investigator, accompanied the blood gasvalues. Each physician received a questionnaire includ-ing the capillary blood gas results and clinical informa-tion. A similar questionnaire for the venous gas was sub-mitted on a separate occasion with at least a 3-monthinterval between submission of the capillary and venousgas results. Each intensivist was asked the following ques-tions; “What is your interpretation of this gas? Would youintubate this child? Would you give NaHCO3? If thispatient was already intubated, would you change theminute ventilation?” To be able to correct for intraindivid-ual variability over time, the intensivists were asked tointerpret each capillary and venous gas value on 2 occa-sions at least 3 months apart. The order of the capillaryand venous gas values were randomly assigned to avoidrecall bias. Physicians were asked whether the source ofthe gas, capillary or venous, influenced their response tothe interpretation or management decisions.

The intraclass correlation coefficient was calculated forthe measured blood gas parameters (pH, PCO2, PO2) using1-way analysis of variance (ANOVA) for repeated mea-

Table 1. Distribution of measured capillary blood gas pH and PCO2.

Abnormal Values No. (%)Parameters No. Range Mean High Low

pH* 78 7.07–7.53 7.41 12 (15.4) 12 (15.4)PCO2

† 78 17–70 34 42 (53.8) 5 (6.4)*High pH defined as greater than or equal to 7.45; low pH defined as less than or equal to 7.35.†High PCO2 defined as greater than or equal to 45 mm Hg; low PCO2 defined as less than orequal to 35 mm Hg.


McGillivray et al


for the question relating to the need for change in venti-lation if the patient was intubated.

When asked if the source of the gas value sample (venousversus capillary) influenced their response to interpreta-tion and management questions, intensivist A and B repliedpositively in 26.9% (21/78) and 11.5% (9/78) of cases. Theyboth believed that in a patient with acidosis, the venous gasvalue might overstate the degree of acidosis.

D I S C U S S I O N

In the well-perfused patient, all parameters for capillary andvenous gases showed excellent intraclass correlation,with the exception of PO2. The excellent correlation observed

with abnormal pH values, the venous-capillary crudeagreements were similar (ie, ≥.63 for gas interpretation,≥.83 for need for bicarbonate, ≥.88 for need for intuba-tion, and ≥.58 for need for changes in ventilation for the2 physicians).

The α values of concordance for intraobserver repro-ducibility on repeated administration of capillary-cap-illary and capillary-venous samples with regard tointerpretation and management decisions are shown inTable 2 for the 78 gas values. Both intensivists showedsimilar capillary-venous concordance. Capillary-venous concordance values were highest for questionsrelating to the need for bicarbonate administration andintubation. Lower levels of concordance were obtained

Table 2. Intraobserver α concordance. Interpretation and clinical decision based on brief clinical information and capillary or venous blood gas values.

Intensivist A Intensivist A Intensivist B Intensivist BCapillary-capillary* Capillary-venous† Capillary-capillary* Capillary-venous†

Parameter (95% CI) (95% CI) (95% CI) (95% CI)

Gas interpretation .75 (.69–.89) .61 (.47–.75) .83 (.78–.88) .48 (.41–.55)Need for bicarbonate .69 (.50–.88) .85 (0.73–.97) .69 (.58–.80) .80 (.72–.88)Need for intubation .80 (.72–.86) .73 (0.64–.82) .96 (.93–.98) .83 (.75–.91)Need for changes in ventilation parameters .59 (.51–.67) .37 (0.27–.47) .56 (.47–.65) .32 (.21–.43)*Intraobserver concordance on repeated administration of the questionnaire with capillary blood gas values.†Intraobserver concordance between administration of the questionnaire with venous versus capillary gas values submitted at least 3 months apart.

Figure 2.Intraclass correlation coefficients (ICC) for venous and capillary measured gas values.


McGillivray et al


Given the low frequency of interventions in ourpatient population, it may be hypothesized that a bloodgas determination in the ED may best be used as ameans of assessing severity of illness or monitoringpatient progress. In a future study, it would be prefer-able to expand the list of questions presented to thephysicians to better determine the utility of the venousgas value in its role as a diagnostic, assessment, andmanagement tool.

The excellent correlation between values obtainedfrom venous and capillary sources observed in thisstudy suggests that the venous gas value provides anacceptable means for assessing severity of illness andresponse to treatment in well-perfused patients. Thenotable intraobserver concordance for gas interpreta-tion, need for bicarbonate, and need for intubationindicate that the venous sampling for blood gas valuesis an appropriate laboratory test to assist the emergencyphysician in making the main management decisions.

R E F E R E N C E S

1. Wetzel R, Tabata B, Rogers C: Hemodynamic monitoring considerations in pediatric criticalcare, in Rogers M (ed) : Textbook of Pediatric Intensive Care, ed 2. Philadelphia: Williams &Wilkins, 1992:622-624.

2. Fuhrman B, Zimmerman J: Pediatric Critical Care, ed 2. St Louis: Mosby–Year Book,1998:138.

3. Barkin R: Pediatric Emergency Medicine, Concepts and Clinical Practice, ed 2. St Louis:Mosby–Year Book, 1997:155.

4. Fleisher G, Ludwig S: Textbook of Pediatric Emergency Medicine, ed 3. Philadelphia:Williams & Wilkins, 1993:1569.

5. Mortensen J: Clinical sequelae from arterial needle puncture, cannulation, and incision.Circulation 1967;35:1118-1123.

6. Hale P, Nattrass M: A comparison of arterial and nonarterialized capillary blood gases indiabetic ketoacidosis. Diabet Med 1988;5:76-78.

7. Brandenburg M, Dire D: Comparison of arterial and venous blood gas values in the initialemergency department evaluation of patients with diabetic ketoacidosis. Ann Emerg Med1998;31:459-465.

8. Harrison AM: Comparison of simultaneously obtained arterial and capillary blood gases inpediatric intensive care unit patients. Crit Care Med 1997;25:1904-1908.

9. Davis H, Beran M, Galant S: Capillary pH and blood gas determination in asthmatic children.J Allergy Clin Immunol 1975;56:33-38.

10. Begin R, Racine T, Roy J: Value of capillary blood gas analysis in the management of acuterespiratory distress. Am Rev Respir Dis 1975;112:879-881.

11. Doig W: Value of arterialized capillary blood gas analysis in lower respiratory tract infectionin childhood. Arch Dis Childh 1971;46:243-246.

12. Gandy G, Grann L, Cunningham N, et al: The validity of pH and pCO2 measurements in capil-lary samples in sick and healthy newborn infants. Pediatrics 1964;34:192-197.

13. MacRae D, Palaavradji D: Comparison between arterial capillary and venous acid base mea-surements in the newborn infant. J Obstet Gynaecol Br Commonw 1966;73:761-765.

14. Courtney S, Weber K, Breake L, et al: Capillary blood gases in the neonate: A reassessmentand review of the literature. Am J Dis Childh 1990;144:168-172.

15. Gambino S, Thiede W: Comparison of pH in human, arterial, venous and capillary blood. AmJ Clin Pathol 1959;32:298-300.

over a large range of values for all parameters is in keep-ing with previous studies in neonates and in smallergroups of children.5,6,9-15

The good crude agreement and concordance observedfor each intensivist on repeated submissions of the cap-illary gas questionnaire indicated good intraobserverreproducibility despite some variability over time. Thisvariability over time must be taken into account in theinterpretation of the capillary-venous crude agreementand concordance. Yet, the crude agreement and concor-dance of capillary-venous pairs remained good for boththe entire group, as well as in the children with abnor-mal pH values. Concordance was particularly high forthe management questions involving need for bicar-bonate administration and need for intubation. Thereliability of the blood gas measurement is very impor-tant in aiding the physician to make these 2 importantmanagement decisions.

These results must be interpreted in light of the fol-lowing. (1) The results pertain only to well-perfusedchildren, although they apply to a large number ofmildly to moderately ill children presenting to an EDwith respiratory or metabolic illnesses. Having estab-lished the utility of venous gas sampling in this popu-lation, the remaining question is whether the venousgas value may be valid in patients who are not well-perfused, using the arterial blood gas as the goldstandard in these patients. (2) It is possible thatsome study patients, particularly those with dis-tinctly abnormal gas values, may have been inade-quately perfused and therefore, incorrectly includedin our sample. Their inclusion would tend to under-estimate the correlation between the capillary andvenous values. The excellent numerical intraclasscorrelation between samples suggests that either thepatient selection was adequate or that venous andcapillary gases values correlate well in the hypoper-fused patient. Despite these interesting results, read-ers should be cautioned that if a venous, capillary, orarterial blood gas value is not in keeping with theclinical picture, the gas value should be redeterminedbefore making management decisions on the basis ofthe questionable gas alone. (3) With regard to theinterpretation and management questions, a largerpool of intensivists would have provided a more pre-cise estimate of the expected concordance betweencapillary and venous gas values. Systematic bias ingas value interpretation and management responseswas reduced by the different training backgroundsof the 2 intensivists.


McGillivray et al


16. Pre-analytical considerations: Specimen collection, calibrations, and controls in: ApprovedGuideline, document c27-A, National Committee for Clinical Laboratory Standards Villanova:National Committee for Clinical Laboratory Standards Publication 1993;13:6.

17. Steiner D, Norman G: Reliability, in Steiner D, Norman G (eds): Health Measurement Scales:A Practical Guide to Their Development and Use. Oxford: Oxford University Press, 1989:79-96.

18. Fleiss J: Reliability of measurement, in Fleiss J (ed): The Design and Analysis of ClinicalExperiments. New York: J Wiley & Sons,1986:1-32.

19. Aitken M: Maximum likelihood estimation of agreement in the constant predictive probabil-ity model, and its relation to Cohen’s kappa. Biometrics 1990;46:293-302.

20. Landis J, Koch GG: The measurement of observer agreement for categorical data.Biometrics 1977;33:159-174.

21. Fluid therapy and medications, in Chameides L, Hazinski MF (eds): Textbook of PediatricAdvanced Life Support, ed 3. Chicago: American Academy of Pediatrics, 1997:2-3.

clinical decisionmaking based on venous versus capillary blood gas values in the well-perfused child

Documents