Reflex testing to define action limits for community-based requests

Abstract

Background

Critical values are required to be phoned 24/7. Other abnormal results fall short of the thresholds used to define critical values and may only be required to be phoned during the day. Community-based requestors prefer not to be contacted unless a result is critical and contacting them requires substantial staff resource. It is common practice to add tests to requests to expedite diagnosis or clarify the significance of a particular result using algorithms.

Methods

We devised algorithms for reflex addition of tests which allowed the differentiation of significantly abnormal results as either critical values or those that only require day phoning.

Results

Algorithms identified 158 out of 309 tests as being critical (51%) over nine months. Reflex addition of serum bicarbonate identified 4% of serum glucose (24.9–37.9 mmol/L) as critical. Use of estimated glomerular filtration rate by reflex addition of serum creatinine identified 68% of serum lithium (1.49–1.99 mmol/L) as critical. Addition of serum potassium, calcium and magnesium identified 21% of serum digoxin (>2.49 nmol/L) as critical and addition of serum potassium and calcium to all samples with serum magnesium (<0.31 mmol/L) identified hypocalcaemia in all cases. The addition of serum creatinine and potassium as markers for rhabdomyolysis-induced acute renal failure did not help in the differentiation of serum creatine kinase > 4999 μ/L.

Conclusions

Use of reflex tests helped inform a phoning system based on the division of results into critical values and non-emergency abnormal values. This avoids disturbing requestors unnecessarily and conserves staff time at night.

Introduction

A critical value is one that is at such variance with normal as to be life-threatening and when reported requires prompt initiation of corrective action and lifesaving therapy. ¹ Poorly implemented reporting of critical values may impair patient care by squandering resource ² and such results may lose their impact if reported excessively. ¹ Therefore, abnormal results may be categorized as those that are:

Critical values requiring to be phoned 24/7;

Significantly abnormal but non-emergency and non-life-threatening – only requiring to be phoned during the day.

Labtests is a community pathology laboratory serving a population of 1.2 million with the furthest collection centre 88 km away. Users of the laboratory are predominantly general practitioners. The last blood sample of the working day is taken at 17:30 and transport to the laboratory may take 1.5 h. This, along with the time taken to receive, process and analyse the samples, means that critical values often present after office hours. Contacting a community requestor out of office hours requires substantial staff resource. It can be difficult to locate requestors ^3,4 because after hours contact details have not been made available, the mobile phone has been switched off or the voicemail option is kept activated, the requestor is on leave or the duty doctor for a practice may refuse to accept the results because the patient is not known to them. Additionally, there may be uncertainty about the significance of the results.

Some test results may indicate that additional useful information about diagnosis or the significance of abnormal results can be gleaned by the addition of further tests. ⁵ These extra tests are often added to a request using algorithms that are usually developed ‘in-house’. Best practice for this activity is not well established or defined. For example, it is recognized that the addition of calcium and potassium when hypomagnesaemia is present, ⁵ the addition of magnesium when hypokalaemia is present, ⁶ and the addition of potassium, magnesium and calcium when digoxin is raised, ⁷ are good practices.

Guidelines on action limits are not evidence-based. ^2,3,8–13 In order to ensure that community-based requestors were only called when necessary with critical values, we devised algorithms for reflex test additions to better define such results and evaluated them prospectively (except for serum lithium which was evaluated retrospectively). Reflex testing algorithms have not previously been used in this way and we evaluated their utility in better defining critical values.

Methods

The literature on critical values was reviewed. ^2,3,8–13 An existing in-house two-tier critical phoning list was reviewed and algorithms for the reflex addition of tests were devised (Tables 1–5) to help better define a critical value. For example, for a range of high-glucose values, we thought it important to determine whether or not the result was associated with systemic acidosis (signifying perhaps ketoacidosis). If serum glucose >37.9 mmol/L, the result was considered sufficiently abnormal to warrant phoning 24/7 no matter what the concurrent acid–base status. Likewise, for a serum lithium concentration of 1.49–1.99 mmol/L, the critical nature of the result was defined by the presence of concomitant kidney impairment. Where appropriate, the algorithms for reflex test addition were approved after consultation with the relevant hospital specialists, and then by the laboratory quality group. The reflex tests were added manually when the test appeared in the critical result list or in the validation list. As samples are delivered to the analysers by a track, the only intervention that is required for reflex test addition is to simply add the request to the laboratory information system. The samples in open-top primary tubes may have been on the track for 30–40 min by the time the extra tests are added. When reflex testing of serum bicarbonate was required, the sample was physically located, mixed and front-loaded onto the platform. Therefore, addition of a reflex test for bicarbonate resulted in an additional five minutes of staff time being used to locate and load the specimen.

Table 1

Glucose: suitable for reflex requests

Intent	To determine the presence/absence of significant acidosis in hyperglycaemia in patients aged >16 y
Number*	∼312,000 in nine months
Intervention	If serum/plasma > 24.9 mmol/L, add serum bicarbonate
Critical result	If glucose >24.9 mmol/L and serum bicarbonate <22 mmol/L
	If glucose >37.9 mmol/L
	If glucose >24.9 mmol/L and newly diabetic
Results	Glucose >24.9 mmol/L on 287 occasions (range 25.8–44.5 mmol/L)
	Glucose >37.9 mmol/L on 15 occasions. Of these: serum bicarbonate <22 mmol/L in three persons and HONK in four persons (aged: 66, 69, 73 and 79 y)
	Values 25.0–37.9: n= 272, bicarbonate <22 mmol/L in 11 (range 8–21 mmol/L)
	New diabetics not included in above: 28
Outcome	54 critical results phoned 24/7. Saving of 233 phone calls
	Labour saved = 233 × 20 min or nine hours per month over nine months

*Total number of requests in the time period

HONK, hyperosmolar non-ketotic hyperglycaemia

Table 2

Lithium: suitable for reflex requests

Intent	To determine the association of high-serum lithium with renal impairment
Number	∼3600 over nine months
Intervention	If serum lithium >1.49 mmol/L, add serum creatinine
Critical result	If lithium >1.49 mmol/L and eGFR < 65 mL/min/1.73 m2
	If lithium >1.99 mmol/L
Results	Lithium >1.49 mmol/L on 39 occasions (range 1.5–3.0 mmol/L)
	Lithium >1.99 mmol/L on 17 occasions
	Values 1.50–1.99: eGFR <65 mL/min/1.73 m2 in 15 (range 18–65)
Outcome	32 critical results phoned 24/7. Saving of seven phone calls
	Labour saved = 7 × 20 min or 0.3 h per month over nine months

eGFR, estimated glomerular filtration rate

Table 3

Digoxin: suitable for reflex requests

Intent	To determine the potential for clinical toxicity of an elevated serum digoxin
Number	∼3600 over nine months
Intervention	If serum digoxin >2.49 nmol/L, add potassium, magnesium and calcium
Critical result	If digoxin >2.49 nmol/L and serum: potassium <3.5 mmol/L or magnesium < 0.75 or calcium >2.55 mmol/L
	If digoxin >3.0 nmol/L
Results	Digoxin >2.49 nmol/L on 28 occasions
	Digoxin >3.0 nmol/L on 14 occasions (range 3.1–4.2 mmol/L)
	Values 2.50–3.00: potassium <3.5 mmol/L in 0; magnesium <0.75 in 0; calcium >2.55 mmol/L in 3
Outcome	17 critical results phoned 24/7. Saving of 11 phone calls
	Labour saved = 11 × 20 min or 0.4 h per month over nine months

Table 4

Magnesium: reflex testing is good practice

Intent	To determine likely association of hypomagnesaemia with hypokalaemia and hypocalcaemia
Number	∼8100
Intervention	If serum magnesium <0.31 mmol/L, add serum potassium and calcium
Critical result	Low calcium and potassium as well
Results	Magnesium <0.31 mmol/L on 13 occasions
	All 13 had low calcium; low potassium in one
Outcome	No savings

Table 5

Creatine kinase: reflex testing is probably good practice

Intent	To determine the potential for significant renal injury associated with rhabdomyolysis
Number	∼28,800 over nine months
Intervention	If serum CK >4999 iu/L, add eGFR and potassium
Critical result	Decrease from previous eGFR or increase from previous potassium by 20%
Results	Serum CK >4999 iu/L on 42 occasions (range 5030–128,000 iu/L)
	eGFR values in this group 82 – >90 mL/min/1.73 m2 except in one patient with eGFR 46 mL/min/1.73 m2 (decrease in eGFR 36%, on statin, acute, required hospitalization)
	Potassium values in this group 3.9–5.6 mmol/L. In no samples was the potassium increase from previous >20%
Outcome	No savings

eGFR, estimated glomerular filtration rate; CK, creatine kinase

After the critical value was established (either by absolute values or by the presence of an abnormal test value and an abnormal reflex test value), the results were phoned by the duty scientist. If it was not possible to locate a clinician who would accept the results, a pathologist was contacted to phone the patient and if they were not contactable, then the police, or ambulance service, was to be contacted.

Results

Data for nine months were reviewed (Tables 1–6). The algorithms identified 158 out of 309 significantly abnormal test results as being critical values (51%) over this period. Reflex request addition of serum bicarbonate identified 4% of serum glucose (24.9–37.9 mmol/L) as critical values; use of estimated glomerular filtration rate (eGFR) identified 68% of serum lithium (1.49–1.99 mmol/L) as critical values (when eGFR < 65 mL/min/1.73 m²); and addition of serum potassium, calcium and magnesium identified 21% of serum digoxin (>2.49 nmol/L) as critical values. All specimens with serum magnesium <0.31 mmol/L had low serum calcium and one had low serum potassium. Addition of serum creatinine and potassium did not help in the definition of serum creatine kinase > 4999 μ/L as a cause of rhabdomyolysis-induced acute renal failure. The estimated savings in time that would have been spent by laboratory staff calling practitioners with results that were not critical values, as defined by these criteria, was 9.3 h per month.

Table 6

Total time saving

Glucose	Saving of 233 phone calls
Lithium	Saving of 7 phone calls
Digoxin	Saving of 11 phone calls
Total calls	Saving of 251 phone calls
	Labour saved = 251 × 20 min or 9.3 h per month

We tracked the follow-up of the 233 patients with glucose results between 24.9 and 37.9 mmol/L which were not defined as critical values: one was seen the same day in the emergency department and discharged the same day and one was seen in outpatients the next day. The mean time to a repeat request for glucose in these patients was 3.6 months, with a maximum of 13 months. In these 233 patients, 226 had glycosylated haemoglobin >75 mmol/mol, indicating chronically poor diabetic control.

Of the 11 digoxin results which were not defined as critical values, three had digoxin therapy stopped, one had serum digoxin repeated the next day, and the remaining seven had a repeat test at 79 days (range 1–350), of which all were normal.

Of the seven lithium results between 1.50 and 1.99 which were not defined as critical values, the follow-up period ranged from 2–70 days and the mean follow-up eGFR increase was 3 mL/min/1.73 m², range 0–9. Of the 17 lithium results which were >1.99 mmol/L, the follow-up period ranged from 2–8 days and the mean follow-up eGFR increase was 8 mL/min/1.73 m², range 0–24. Of the 15 lithium results 1.5–1.99 mmol/L with an eGFR < 65 mL/min/1.73 m², the follow-up period ranged from three to 125 days and the mean follow-up eGFR increase was 11 mL/min/1.73 m², range 6–30.

Discussion

We developed a system where the reflex addition of tests has improved our ability to identify critical values, through the generation of a profile of test results, which we believe allows a more useful definition of what constitutes a critical value. This has helped inform a critical value phoning system for glucose, lithium and digoxin, and avoids disturbing requestors out of hours with abnormal results. It saves significant staff resources – allowing them to focus on other important laboratory duties, especially at night, when staff numbers are low. Staff regarded the reflex tests as helpful to critical value phoning. An audit of critical values found that a call about outpatient results took on average 14 min, and at night was far longer than during the day. ⁸ We believe that the study confirms that the addition of potassium and calcium is good practice with low magnesium. It has not proved helpful in further defining critical values of creatine kinase. Patients with previously known chronic muscle disorders had eGFR > 90 mL/min/1.73 m² due to low serum creatinine concentrations. However, addition of eGFR and potassium is probably good practice in this context. ¹⁴

The manual system of reflex testing is dependent on staff adhering to departmental policy. In an attempt to remove any variability in applying laboratory policy, we propose to automate the reflex addition of these tests. The delay in waiting for reflex test results means that a general practitioner is called about 15 min later than they would have been if the result was phoned as soon as it was available. We estimate that this time will be halved when the reflex test additions are automated. There are a number of other areas where we are considering reflex testing algorithms. For example, the addition of creatinine and potassium to requests when serum urea >20 mmol/L and the addition of serum magnesium if serum calcium is low.

The critical values we use for lithium are in keeping with other Australasian community laboratories. The cut-off for lithium toxicity is ill-defined. ^15,16 Lithium directly affects kidney function and toxicity may be linked to kidney function because it is eliminated through the kidney and occurs more in the elderly in whom the eGFR is lower. We believe that use of eGFR is a better marker than the use of serum creatinine alone in defining the deterioration of kidney function because of concomitant loss of glomerular function with muscle mass with age. The follow-up data showed that on repeat testing, serum lithium was lower and eGFR higher in the majority of patients with raised serum lithium.

The follow-up audit of patients with abnormal results for glucose and digoxin showed that in our small sample, no untoward events occurred. It may be considered that some of our critical values are at the extreme end of being acceptable. However, it should be remembered that these have been reviewed by and agreed upon by local clinical experts. Results not phoned immediately (because reflex testing put them into the ‘non-life-threatening’ rather than ‘critical’ category) were phoned the next day. In conclusion, we have shown that the use of reflex tests with significantly abnormal results can be used to inform a two-tier phoning system for such results that avoids disturbing requestors unnecessarily and conserves staff time at night.

Declarations

Competing interests: None.

Funding: No funding was required.

Ethical approval: Not applicable.

Guarantor: JB.

Contributorship: All authors were involved in the conception, design and implementation of the study. Laboratory staff were supervised by LA and L-JR. Data were analysed by JB and GS. All authors reviewed and edited the manuscript and approved the final version.

Acknowledgements: We thank Myles Wright and Brent Glanville for data capture and Lorraine Elliot for excellent secretarial support.