Assessment of a Critical Limit Protocol for Point-of-Care Glucose Testing

Transcription

1 CLINICAL CHEMISTRY Original Article Assessment of a Critical Limit Protocol for Point-of-Care Glucose Testing GIFFORD LUM, MD A critical limit protocol requiring that all point of care glucose meter readings >22.2 mmol/l (400 mg/dl) and <2.2 mmol/l (40 mg/dl) be immediately confirmed by the laboratory was assessed. A total of 193 (2%) of 9,523 glucose meter determinations (63 patients) were >22.2 or <2.2 mmol/l. One hundred twenty-two (63%) of critically high and low glucose readings were followed up, and 71 (37%) results were not. Seventy-seven percent (55 of 71) of results without follow up were in patients with multiple glucose meter/central lab comparisons, suggesting that users may have thought it unnecessary to confirm such results. Split sample quality control specimens showed good correlation (r = 0.927) between glucose meter and central lab results, whereas correlation for follow-up glucose results was poorer (r = 0.793), perhaps reflecting time delay in obtaining a lab sample. For follow-up results, only 18% of high/low critical limit glucose meter readings were confirmed by drawing a lab specimen within 10 minutes. Fifty-eight percent were in 17 patients with multiple previous glucose meter readings, suggesting that users may have thought it less urgent to confirm a sequence of such results. Eleven follow-up results (9%) showed a >50% discordance between glucose meter/central lab with three (27%) glucose meter errors, emphasizing the need to confirm critically high/low glucose meter results to avoid potential errors. The critical limit protocol now requires that only the initial critically high/low glucose meter reading be confirmed by the lab and that these patients now be followed with lab values until glucose levels are between mmol/l ( mg/dl) before the glucose meter can again be used. (Key words: Point-of-care; Glucose testing; Critical high/low limit; Hyperglycemia; Hypoglycemia) Am J Clin Pathol 1996; 106: Point-of-care glucose testing in some sites has become an integral part of the management of the diabetic patient. Guidelines for point-of-care glucose testing should be established to optimize its use, including the establishment of high and low critical limits designed to provide urgent physician notification of patients who may need immediate diagnostic evaluation and possible therapy for lifethreatening emergencies. Because of the acute deleterious effects of hyperglycemia, which may include acidosis, coma and death, and of hypoglycemia, which may include confusion, dizziness, blurred vision, deterioration in cognitive function, coma and death, medical centers have established critical limits for high and low glucose levels that are used for urgent physician notification. Generally, the critical limit for hyperglycemia varies between mmol/l (110-1,000 mg/dl) and for hypoglycemia between mmol/l (30-70 mg/dl). 1 ' 2 A recent national survey of 92 medical facilities, including 20 trauma centers, found a mean critical limit for hyperglycemia of 26.9 mmol/l (484 mg/dl) (standard deviation [SD] of 8.0 mmol/l [144 mg/dl]) and for hypoglycemia of 2.6 mmol/l (46 mg/dl) (SD of 0.4 mmol/l [7 mg/dl]). 1 At this medical center, for point-of-care glucose meters, the critical limit for hyperglycemia was established as any glucose level > 22.2 mmol/l (400 mg/dl) and for hypoglycemia as any glucose level < 2.2 mmol/l (40 mg/dl). The critical limit glucose protocol requires that all glucose readings >22.2 and <2.2 mmol/l be verified by submitting a STAT specimen to the central clinical laboratory for confirmation by a reference glucose method. To assess the effectiveness of our present critical limit protocol for glucose meters, I decided to study all critically high and low glucose meter determinations for a period of 3 months. MATERIALS AND METHODS From the Pathology and Laboratory Medicine Service, Brockton/ West Roxbury Veteran Affairs Medical Center, Boston, Massachusetts and Harvard Medical School, Boston, Massachusetts. Manuscript received March 12, 1996; revision accepted May 17, Address reprint requests to Dr. Lum: Pathology and Laboratory Medicine Service, Brockton/West Roxbury Veteran Affairs Medical Center VFW Parkway, Boston, MA Patient Samples and Analytical Methods Patient blood samples for point-of-care glucose testing were obtained by nursing staff by fingerprick. Analysis for blood glucose was performed on the ward using a Satellite G (Sat G) Blood Glucose Monitoring System (Medisense, Waltham, MA) and sensor electrode test strips for 390

2 LUM 391 A Glucose Criti Limit Protocol blood glucose, a drug reagent technology using an electrochemical detection technique measuring the electrical current generated from the enzymatic cleavage of glucose. 3 Blood glucose results were reported without correcting the result for the effect of plasma volume and are linear between mmol/l. Patient samples for centralized laboratory testing were obtained by venipuncture in heparinized vacutainer blood collection tubes, mixed and submitted to the central laboratory for glucose determination. Plasma glucose levels were determined on a Beckman Synchron CX-7 automated chemistry analyzer (Beckman Instruments, Brea, CA) using a glucose oxidase method with an oxygen electrode 4 and are linear between mmol/l. Reference Range and Definition of Critical Limits for Hyper and Hypoglycemia The reference range for whole blood glucose levels was mmol/l (65-95 mg/dl) and for plasma glucose levels, mmol/l ( mg/dl). The critical limit for hyperglycemia for the glucose meter was defined as any blood glucose level > 22.2 mmol/l (400 mg/ dl) and for hypoglycemia as any level <2.2 mmol/l (40 mg/dl). Study Subjects This study, involving the use of medical data from human subjects, fell into the category of exempt research in accordance with the federal policy for the protection of human subjects (38 CFR Part 16). All patients studied were adults (> 18 years of age); no pediatric patients were included in this study. Quality Control Program for the Satellite G Glucose Meter A total of 200 trained and certified registered nurses are responsible for operating 21 Sat G meters on 17 in house wards, including ambulatory care sites and emergency room. An ancillary glucose committee (laboratory, nursing, medical staff) is responsible for the overall direction and quality control of the glucose meter program. All Satellite G glucose meter users must be trained and certified before they can use the glucose meter in a clinical situation. Recertification of operators is required on a semi-annual basis or when there is evidence that an operator is not adhering to established medical center guidelines. Quality control material (high and low control) is performed once per day on each day of glucose meter use for patient results. Once a week, the glucose level of an identical heparinized plasma split sample is analyzed on the glucose meter and by the clinical laboratory. Proficiency samples (Ancillary Whole Blood Glucose survey, College of American Pathologists [CAP]) 5 are done 3 times per year for each glucose meter. Critical Limit for Hyperglycemia and Hypoglycemia: Protocol and Assessment The critical limit protocol at this facility requires that all critically high and low glucose meter readings be confirmed by submitting a confirmatory STAT glucose specimen to the central lab, regardless of whether the same patient had multiple previous glucose meter results. All glucose meter results are entered directly into the computer by the user, with notation of time, and for critically high and low glucose readings, the time of drawing a confirmatory laboratory specimen is noted. The critical limit protocol was assessed using the following criteria: (1) glucose meter results with/without follow up; (2) time interval between glucose meter reading and drawing lab specimen; (3) concordance distribution for glucose meter readings and central lab results; (4) review of results and of patient data for glucose meter/ central laboratory follow-up results with >50% discordance; and (5) assessment of patient outcome data for results with no lab follow up. RESULTS Quality Control Program and Satellite G Glucose Meter Data Over a 6-month period, the coefficient of variation (CV) for the low control (2.6 mmol/l, 47 mg/dl) was 7.8% and for the high control (19.4 mmol/l, 350 mg/ dl), the CV was 6.7%. For 200 split samples, the mean whole blood glucose was 5.3 mmol/l (96 mg/dl) for the Satellite G vs a mean plasma glucose of 6.3 mmol/l (114 mg/dl) for the central laboratory (mean difference of 16%, range mmol/l, mg/dl). The correlation coefficient r was 0.927, and linear regression was as follows: Satellite G result (y) = 0.87 X laboratory result For the last three mailings of the Ancillary Whole Blood glucose surveys of the College of American Pathologists WBC A,B,C (1995), 5 only one unacceptable result was found for the 21 glucose meters used at this facility. Critical High/Low Glucose meter Readings During the 3-month period of study, there were a total of 9,523 glucose meter determinations. Figure 1 shows Vol. 1 No. 3

3 392 CLINICAL CHEMISTRY Original Article 193 results inpatients Only 18% of specimens were drawn for lab confirmation within 10 minutes of a critically high and low glucose meter reading, with an additional 20% and 16% within 20 and 30 minutes, respectively. Hence, slightly more than half of the laboratory comparison specimens were drawn within 30 minutes of the initial critical limit high or low glucose meter reading, but in some instances, nurses waited as long as 2 hours to confirm the critical limit result. Seventy-one (58%, 71 of 122) of the followup results were in 17 patients with multiple glucose meter readings, but in only 4 of these patients was the confirmatory laboratory specimen drawn within 10 minutes. >22.2 mmow. 168(51) <22 mmom. 25(12) Concordance Between Glucose Meter Readings and Central Laboratory FQkNV Up-106 (42) NO FOllOWUp-62 (17) Previous- 50(8) Clerical error-5(1) No previous-9 (8) UP-16 (9) NO FOllOWUp-9 (5) Prevlous- 5(2) Clerical error-2(d No previous-2 (2) FIG. I. Flow chart for critical high/low glucose meter readings. the data for the 193 glucose meter results (2.0% of total), which were >22.2 or <2.22 mmol/l. For glucose meter results more than the critical limit of 22.2 mmol/l, operators followed up the high result in 106 instances (42 patients) and did not follow up 62 results (17 patients). For results with no follow up, 81 % (50 of 62) were in eight patients who had previous glucose meter results (consistent with hyperglycemia), 5% (3 of 62) were clerical errors in 1 patient, and 15% (9 of 62) were in eight patients with no previous glucose meter/ laboratory comparisons. Thus, the majority of results that were not followed up were in patients who already had a sequence of high glucose results confirming hyperglycemia. For the 25 glucose meter readings <2.2 mmol/l (12 patients), there was follow up for 16 results (9 patients) and no follow up for 9 results (5 patients). For the results with no follow up, 56% (5 of 9) were in two patients with previous glucose meter/lab results that confirmed hypoglycemia, 22% (2 of 9) were clerical errors in one patient, and 22% (2 of 9) were in two patients with no previous glucose meter/laboratory comparisons. As previously noted, a correlation coefficient (r) of was found for split sample quality control specimens, whereas a r = was found between the glucose meter and central lab glucose results for the 122 follow-up specimens. It is possible that point-of-care glucose meters/central lab comparisons are poor because the initial result is not followed up in a timely manner. Quality control split sample results indicate that when performed with little or no time difference, the glucose meter results have a good correlation with the central lab. Figure 3 shows the concordance distribution for the glucose meter readings versus the central lab for split sample (quality control procedure) results and for follow-up specimens. Lower glucose meter versus laboratory results were designated as negative, whereas higher glucose meter versus laboratory results were designated as positive. For split sample quality control comparisons, 88% of glucose meter results are lower and none of the glucose meter results vary by >50% compared to the central lab. In contrast, for the 122 glucose meter results with follow up, 58% were lower and 9% showed more than 50% variation compared to the central lab. Results Exceeding 50% Concordance Between the Glucose Meter/Laboratory (Table 1) Glucose meter results were erroneous for two patients with critically high glucose meter readings. Patients 7-9 showed relatively large percentage differences between 25 Percent Time Between Critical High/Low Glucose Meter Reading and Laboratory Specimen (Fig. 2) Minutes FIG. 2. Time interval between glucose meter results and time that laboratory specimen was drawn for specimens with follow up. A.J.C.P.-September 1996

4 LUM 393 A Glucose Critical Limit Protocol % of results three died within a month, two with acute myocardial infarction, and one with a lung malignancy (poor outcome not directly related to the hyperglycemia). Two critically low glucose meter readings in two patients were not followed up. Both of these patients were treated for hypoglycemia and subsequently followed by central laboratory glucose results without adverse outcomes. DISCUSSION * ^V" ^V> -yv 1 ^V* ^P & «?V" «?V> ^f ^ Jf -?\P >* J? >* J? <b* N* 1? <!? ** **,** H Split specimen I % Concordance Follow Up specimens FIG. 3. Concordance distribution (10 percentile intervals) for glucose meter versus central laboratory glucose results. the glucose meter reading and the laboratory attributable to treatment for hypoglycemia initiated before the laboratory specimen was drawn for analysis. The hypoglycemic result in patient 10 could not be confirmed by the laboratory, and the glucose meter result was erroneous. Thus, overall, for these 11 discordant results, 3 (27%) were found to be glucose meter errors. Outcome for Patients With No Glucose Meter/Central Laboratory Follow up (Table 2) Eight hyperglycemic patients were appropriately treated with subsequent lab monitoring of glucose levels; The glucose results of the Satellite G glucose meter (whole blood) and the central laboratory's reference glucose method (plasma) are not directly interchangeable. Typically, the glucose level in plasma will be higher than the glucose level in whole blood cells because the average water content of plasma is approximately 93%, whereas it is only about 71 % in red blood cells. 6 A mathematically derived factor of 1.12, assuming a hematocrit of 45% and a red cell:plasma water ratio of 0.76, 7 may be used to convert whole blood glucose to plasma glucose. However, neither of these assumptions may be true in many patients especially in those with dehydration, hyperosmolality, and hyperglycemia. 7 Despite this difference in whole blood/plasma glucose values, for split samples of glucose meter/central laboratory comparisons, a significant correlation (r = 0.927) was found. A less significant correlation (r = 0.793) was found between glucose meter and central lab follow-up glucose results. The relatively poor comparison between glucose meter and central lab in situations where there TABLE 1. SUMMARY OF GLUCOSE METER AND LABORATORY GLUCOSE RESULTS WITH >50% DISCORDANCE Patient Satellite G Laboratory (mmolll) Time Difference No. (mmolll) [% difference] (min) Diagnosis Outcome (-72) 5.8 (+82) 7.2 (+70) 2.2 (-83) 3.1 (-55) 3.3 (-200) 7.4 (-363) 6.4 (-276) 3.8 (-138) 13.5 (-864) 22.6 (-1955) Pancreatic carcinoma Diabetes, end stage renal disease Angina, CABG, no diabetic history Brittle diabetes, hypoglycemic episodes Diabetes with ketoacidosis Insulin-dependent diabetes Diabetes, hypoglycemia thought to be secondary to poor p.o. intake and high-dose glyburide Insulin-dependent diabetes Diabetes, markedly hypoglycemic Metastatic renal cell carcinoma, steroid-induced worsening of diabetes Died 4 days after high glucose noted Patient not treated, no glucose > 11.1 mmol/l, glucose meter error Patient not treated, glucose meter error Patient treated for hypoglycemia Patient treated for hypoglycemia Patient treated for hypoglycemia Glyburide stopped, patient treated with glucose before laboratory specimen drawn Patient treated for low glucose before laboratory specimen drawn Patient treated with dextrose bolus at same time laboratory specimen drawn Hypoglycemia not confirmed by laboratory, no laboratory result < 12.8 mmol/l, patient not treated for hypoglycemia, glucose meter error CABG = coronary artery bypass graft. Vol. 106-No. 3

5 394 CLINICAL CHEMISTRY Original Article TABLE 2. SUMMARY OF OUTCOME IN PATIENTS WITH NO GLUCOSE METER FOLLOW-UP Patient No. Satellite G (mmoi/l) Diagnosis Outcome > <1.1 <1.1 Diabetes, quadriplegic, prostate carcinoma Spinal cord injury patient with diabetes and non-q wave myocardial infarct X 2 Diabetes with below knee amputation and osteomyelitis Non-small cell carcinoma of the lung Diabetic with third-degree burn Spinal cord injury patient with insulin-dependent diabetes Poorly controlled insulindependent diabetes Diabetes and osteomyelitis 84-year-old with cognitive decline, on ADA diet with? hypoglycemia Cirrhosis of liver and hepatitis A, B, and C, alcoholic and drug dependence, diabetes Patient died 20 days after high glucose noted, transmural myocardial infarct, patient treated with insulin but died 22 days after high glucose noted (acute subendocardial MI) and followed by laboratory results but died 2 days after high glucose noted and followed by laboratory and then followed by laboratory and followed by laboratory Patient changed to regular diet, treated with glucose, no evidence of glucose intolerance Patient treated for hypoglycemia and followed by laboratory results ADA = American Dietetic Association; Ml = myocardial infarction. were delays in central lab analysis may well be related to the time factor because, if used as intended (little or no time differential), the glucose meter and central lab glucose results have good correlation. Another factor that may account for the poor correlation for glucose meter/ central lab results for follow-up specimens would be initiation of treatment for hyper or hypoglycemia. There is strong suggestive support for this explanation because there is no correlation (r = 0.107) between time interval and glucose differences (initial glucose meter and central lab glucose result). Approximately one third of critically high and low glucose meter results (37%, 71 of 193) were not followed up, a strikingfindingof this study. Seventy-seven percent (55 of 71) of these results lacking follow up were in patients who already had a sequence of high or low glucose results, suggesting that users may have felt it unnecessary to reconfirm the abnormal glucose reading with a separate lab specimen. Many studies have focused on the acceptable turnaround time between actual test ordering and receipt of results by the clinical team. An acceptable rapid response time during critical emergencies of 5 minutes or less from the ordering of the test to test result receipt has been recently recommended. 8 A recent study found a mean turnaround time of 8.5 minutes for ward glucose meter determinations from specimen collection to test completion. 9 We found that only 18% of specimens were actually drawn for lab confirmation within 10 minutes, a less than optimal response. Fifty-eight percent of the follow-up results were in 17 patients who had multiple glucose meter readings and were known to be hyper or hypoglycemic; hence, this time delay may reflect less urgency to confirm a sequence of such glucose results on the part of users. Other reasons cited for delay included lack of time ("too busy"), waiting for phlebotomist to draw specimen, and results of glucose meter considered reliable for therapeutic action. Our data found that 9% of glucose meter results showed a >50% variation compared to the central lab for those results with follow up, indicating that it is important to confirm critically high and low glucose meter results because approximately 25% of the discordant results were glucose meter errors. No quality control split sample comparisons between the glucose meter and central lab showed a >50% variation. For glucose meter results that were not followed up (patients subsequently followed with central lab glucose A.J.C.P.- September 1996

6 LUM 395 A Glucose Criti Limit Protocol results), outcome data indicated that this course of action resulted in no adverse consequences directly related to hyper or hypoglycemia in 10 such patients. Patients with initially critically high/low glucose results were treated appropriately and then followed by the central lab without adverse outcome. An excellent level of precision for glucose meter quality assurance has been achieved by our nurse users (CVs were 7.8% and 6.7% at low (2.6 mmol/l) and at high (19.4 mmol/l) glucose levels, respectively). A recent consensus conference recommended performance goals for glucose meter testing of a total error (analytical plus user) of < 10% at glucose levels between mmol/ L ( mg/dl) 100% oftime,' oracvof <4%. 7 " A more recent consensus conference recommended a target performance for future glucose meters of an analytical error of 5%.' 2 These target goals clearly may not be achievable in the real world because a recent CAP survey found CVs for glucose meters ranged from 4.2% 35.3%, 5 with CVs for the Medisense Satellite G of 6.8% to 9.7%. 5 Did point-of-care glucose testing result in a decrease in laboratory workload? I compared the total number of glucose levels performed by the central lab before and after the introduction of the glucose meter at this facility. There were 20,406 laboratory glucose measurements for a randomly selected preglucose meter quarter compared to 24,028 glucose determinations for the quarter studied, a 18% increase, (the hospital census and patient population mix were comparable for the 2 quarters). Thus, glucose meters did not result in a decrease in lab glucose workload, a finding similar to those of others who found an increased annual lab glucose workload of approximately 10% after the introduction of glucose meters. 13 The medical center protocol for the critical limit high/ low for glucose meters has been modified to incorporate the findings in this study. These findings included relatively poor compliance with the requirement to follow up all glucose meter results in every instance of a critically high and low glucose meter reading, delay between noting an abnormal glucose reading and drawing a confirmatory lab specimen, the need to confirm glucose meter results because of potential erroneous glucose meter results and no adverse outcome directly related to hyper or hypoglycemia for patients whose glucose meter results were subsequently followed with central lab results. The protocol now requires that critically high/low glucose meter readings be confirmed by drawing a STAT lab specimen but that such patients now be followed with laboratory glucose results until the plasma glucose level is within the range of mmol/l ( mg/ dl), selected for ease of recall of conventional concentration units, at which time the glucose meter can again be used. This protocol should result in reducing the confusion among users regarding the need to confirm repetitive critically high/low glucose meter readings in the same patient and result in reducing the potential for glucose meter errors. REFERENCES 1. Kost GJ. Critical limits for urgent clinician notification at US medical centers. JAMA 1990;263: KostGJ. Using critical limitsto improve patient outcome. Medical Laboratory Observer 1993;25: Marks V, Dawson A. Rapid stick method for determining bloodglucose concentration. BMJ 1965:i: Kadish AH, Litle RL, Sternberg JC. A new and rapid method for determination of glucose by measurement of rate of oxygen consumption. Clin Chem 1968; 14: College of American Pathologists Whole Blood Glucose. Multiple Site Series 1 Survey, 1995 Set WBG A,B,C, Northfield, Illinois. 6. Howanitz PJ, Howanitz JH, Henry JB. Carbohydrates. In: Henry JB, ed. Clinical Diagnosis and Management by Laboratory Methods, ed 18. Philadelphia: WBSaunders. 199l.pp Rainey PM, Jatlow P. Monitoring blood glucose meters [Editorial]. Am J Clin Pathol 1995; 103: Kost GJ. Guidelines for point-of-care testing; Improving patient outcomes. Am J Clin Pathol 1995; 104(Suppl):SI 11-S Winkelman JW, Wybenga DR. Tanasijevic MJ. The fiscal consequences of central vs distributed testing of glucose. Clin Chem 1994;40: American Diabetes Association. Consensus statement on selfmonitoring of blood glucose. Diabetes Care 1987; 10: Fraser CG, Petersen PH. Desirable Performance Standards for Imprecision and Bias in Alternate Sites: The Views of Laboratory Professionals. Arch Pathol Lab Med 1995; 119: American Diabetes Association. Self-monitoring of blood glucose. Diabetes Care 1994; 17: Lee-Lewandrowski E, Laposata M, Eschenbach K, et al. Utilization and cost analysis of bedside capillary glucose testing in a large teaching hospital: Implications for managing point-of-care testing. Am J Med 1994;97: No. 3