To test or to not test: A retrospective cross-sectional study on potentially inappropriate use of pathology testing in South Australian hospitals

Abstract

Objectives

To measure rates of potentially inappropriate pathology testing in the hospital setting.

Methods

Retrospective cross-sectional study in hospital setting from July 2021 to December 2021. We examined 3 potentially inappropriate uses: overordering, selection errors, and unnecessary repeat testing. Overordering included vitamin D and lipids (rarely required in acute hospital care). Selection error was the ratio of iron studies to standalone ferritin requests. Unnecessary repeats included any repeat vitamin D, lipids, iron, or ferritin in an episode of care or C-reactive protein (CRP) repeated within 3 days and N-terminal pro–brain natriuretic peptide (NT-proBNP) within 7 days and repeated previously abnormal CRP and NT-proBNP tests. Costs of inappropriate tests were estimated using the Australian Medicare Benefits Schedules.

Results

Among 55,904 test requests, 15% (n = 8120) were potentially inappropriate. Vitamin D was frequently ordered (n = 4498), as were lipids (n = 2872). Ratio of iron studies to standalone ferritin was 36. Of 19,233 repeat CRPs, 36% (n = 6947) were within 3 days and 62% (n = 179) of repeat NT-proBNPs were within 7 days of the first test. For initially abnormal tests, 89% of CRPs and 97% of NT-proBNPs remained abnormal. Inappropriate test costs accounted for 12% to 30% of costs.

Conclusions

Frequent potential inappropriate use and selection of pathology tests was observed in South Australian hospitals.

Key Points.

• Pathology informatics can provide a rich source of test usage data to support quality improvement efforts in the hospital setting.

• Inappropriate pathology testing can occur due to use of an unnecessary test, selection errors, or retesting within a time interval where results are unlikely to contribute further information.

• Costs associated with potentially inappropriate testing are significant, ranging from 12% to 30% of total costs.

INTRODUCTION

Laboratory test results inform 60% to 70% of medical decisions, playing a vital role in diagnostics and disease management.¹ The Australian pathology workforce conducts close to 500 million laboratory-based pathology tests annually, with more than 11 million Australians having at least 1 test every year.² In the context of Australian health care, expenditure relating to the Medicare Benefits Schedule (MBS) for pathology services (nonhospital) equated to 4.1 billion Australian dollars in 2022, comprising 15% of total MBS item spending.³

Investigation of the quality use of pathology is an emerging and growing research focus, guided by pathology informatics.^4,5 While the effective use of pathology provides essential information for the treatment of patients, inappropriate use of pathology can result in missed diagnoses, unnecessary downstream interventions, excess blood draw, and increased costs to the health care system.^6,7 Despite the growing interest in understanding pathology utilization,⁸ the frequency of inappropriate testing has not been widely studied.

Several approaches can be taken to identify what constitutes inappropriate testing. Some such examples can include the following types of inappropriate testing:

• Unnecessary tests are those performed without appropriate clinical indication.

• Inappropriate selection of tests occurs when, for example, a second line test is ordered over a first line test, or panels of tests are ordered when a single test will suffice in diagnosing the patient.

• Unnecessary repeat testing is another form of inappropriate testing.⁹ While the first test may be necessary, repeating the test may present no new information to support patient care and thus is inappropriate. Unnecessary repeat testing is costly to the health care system and can result in patient harm.¹⁰ Although there are clinical guidelines with recommended retest periods,¹¹ it is unknown how well these are followed, especially in the hospital setting.

The aim of this study was to examine the extent of unnecessary test requests, the selection of inappropriate tests, and unnecessary repeat tests, with the use of pathology administration data. The focus for the analysis was requests for vitamin D, lipid profiles (consisting of cholesterol, triglycerides, low-density lipoprotein cholesterol [LDL], and high-density lipoprotein cholesterol [HDL]), ferritin tests as part of iron studies (which collectively include transferrin, iron, ferritin) and standalone ferritin, C-reactive protein (CRP), and N-terminal pro–brain natriuretic peptide (NT-proBNP). The rationale for selecting the tests was as follows.

Vitamin D tests are not usually indicated in the hospital care setting,¹² and while they may be appropriate in some conditions, such as malabsorption or commencing treatment for osteoporosis,¹³ it is generally not a test for the tertiary care setting.¹²

In an acute coronary presentation, an LDL cholesterol assessment (calculated from HDL cholesterol, triglycerides, and total cholesterol) is appropriate within 24 hours of when symptoms first manifested,¹⁴ but repeat testing should not occur within at least 6 weeks.¹⁵ In addition, the current Australian guidelines on management of acute coronary syndrome recommend that statin therapy should commence regardless of baseline LDL level in this patient group.¹⁶

Triglyceride measurements in the context of acute presentations such as acute pancreatitis or diabetic ketoacidosis are appropriate and need to be regularly monitored.¹⁷ However, triglycerides should ideally be ordered as a single test.

A single ferritin measurement is usually sufficient in the assessment of a patient’s iron status¹⁸; however, iron studies are typically ordered instead of ferritin alone.¹⁹

C-reactive protein is a nonspecific marker of inflammation²⁰ and often a high-volume test in hospital care.²¹ However, given the kinetics of CRP, elevated results are unlikely to change significantly within 3 days of the first test,²² and even with abnormal results a repeat should only be done after at least 2 weeks.²³

Elevated NT-proBNP is a first-line indicator of heart failure, and a single measurement is considered sufficient in a hospital presentation.^11,24,25

METHODS

Study Design

We used a retrospective cross-sectional study. The study period was from July 11, 2021, to December 30, 2021.

Setting

The study was set in 2 major teaching hospitals in South Australia, with a combined total of 1100 beds.

Participants

Included patients were those who presented for an emergency department (ED) or inpatient encounter and received a target test during their visit. As data on patient length of stay were unavailable, an episode of care was defined as 30 days from the first instance of a test, known as the index test.

Exposures

The tests assessed were vitamin D, lipid profile (cholesterol, triglycerides, LDL cholesterol, and HDL cholesterol), ferritin ordered as part of iron studies (which includes transferrin, iron, and ferritin tests) and standalone ferritin, CRP, and NT-proBNP.

Data Sources

The data were obtained from the SA Pathology laboratory database. SA Pathology is the sole provider of pathology services to South Australian public hospitals.

Inclusion and Exclusion Criteria

Duplicate records and records with missing data were excluded; the filtering process to derive the final data files is illustrated in Supplementary Figure 1 (all supplementary material is available at American Journal of Clinical Pathology online). All analyses were conducted using SAS 9.4 (SAS Institute) and Excel (Microsoft).

Defining Potentially Unnecessary Tests

For all tests, we defined repeat tests as a subsequent test within 30 days (episode of care) of the index test, at the same hospital.

Potentially unnecessary tests were investigated as follows:

1. Vitamin D: any ordering of vitamin D tests and the proportion of repeat tests within an episode of care. As we did not have access to diagnostic data, we could not determine under which indications an initial vitamin D test was appropriate, and as such, we investigate repeated vitamin D tests for inappropriate ordering behavior.

2. Lipid tests: any ordering of lipid tests and the proportion of repeat tests within an episode of care (30 days).

3. Iron and ferritin tests: ratio of ferritin tests ordered as part of iron studies compared to standalone ferritin tests. The proportion of repeats within an episode of care for both iron studies and standalone ferritin was also reported.

4. CRP: proportion of all repeat tests within 3 days of the index test in an episode of care.

5. NT-proBNP: proportion of all repeat tests within 7 days of the index test in an episode of care.

We further stratified the first repeat test by the index test result, observing what proportion of initially abnormal tests remained abnormal in the first repeat for lipids, CRP, and NT-proBNP.

The proportion of all repeat tests was calculated by dividing the number of repeat tests within each episode of care (30 days) by the total number of tests within each episode of care.

To get an indicative cost associated with the tests, we used the Australian MBS.²⁶ The cost identified here is the full fee of the test, as an estimate. Public patients in public hospitals receive care that is funded by the state government.²⁷ We did not have access to the true costs associated with these tests and as such used the MBS schedule full item fee as an estimate of the cost. The fees listed in MBS are provided as a reasonable average based on test complexity, time, and technical expertise.²⁸

We used 2 definitions to provide an estimate of the overall cost associated with inappropriate tests. Our primary definition was the most conservative: cost of repeat tests for any of the tests repeated prior to the recommended repeat testing timeframe. As a sensitivity analysis, we also included the costs of all vitamin D and lipid profile testing in addition to the repeat test costs as potentially unnecessary.

Statistical Analysis

Descriptive statistics are reported as frequencies and proportions. The statistical strength of associations between index and repeat test results was determined by a χ² test or Fisher test, as appropriate. The significance level was set at P < .05.

Ethics

Ethics approval for the study was obtained from the Central Adelaide Local Health Network Ethics Committee and the University of South Australia Human Research Ethics Committee.

RESULTS

During the study period, there were 55,904 requests for the selected tests. Table 1 shows the overall number of tests and number of repeat tests for each test within each episode of care. The proportion of potentially inappropriate repeat tests with regard to the index test is shown in Table 2. Results for potentially inappropriate repeat tests were 62% for NT-proBNP, 36% for CRP, and 7% for iron studies. A substantially higher volume of iron studies was ordered, compared to a standalone ferritin, with the ratio of iron studies to standalone ferritin tests being 36.

TABLE 1.

Number of Tests and Repeats During an Episode of Care Over the Study Period in South Australian Hospitals

Test	No. of episodes of care	No. (%)of repeat tests within each episode of carea	No. of episodes of care with more than 1 test
Vitamin D	4082	413/4082 (10)	394
Lipid studies	2676	196/2676 (7)	183
Iron studies (ferritin)	5108	367/5108 (7)	342
Standalone ferritin	132	16/132 (12)	9
CRP	20,901	19,233/20,901 (92)	7519
NT-proBNP	2493	287/2493 (12)	244
Total	35,392
CRP and NT-proBNP (repeat before 3 days and 7 days)	No. of repeat tests in an episode of care	No. (%) of repeat tests	No. of episodes of care with a repeat test
CRP (within 3 days of incident)	19,233	6947/19,233 (36)	5137
NT-proBNP (within 7 days of incident)	287	179/287 (62)	153

CRP, C-reactive protein; NT-proBNP, N-terminal pro–brain natriuretic peptide.

^aSome patients will have more than 1 repeat test in an episode of care (30 days).

TABLE 2.

Normal and Abnormal Index and First Repeat Tests: Strength of Association Between an Abnormal Index Test Result and the First Repeat Test

Test and repeat interval	No. of abnormal repeats where index test was abnormal	No. of abnormal repeats where index test was normal	Proportion of abnormal repeat tests, %	P value
Lipid studies (within 30 days)	77	21	79	<.0001
CRP (within 3 days)	3835	458	89	<.0001
NT-proBNP (within 7 days)	140	4	97	<.0001

CRP, C-reactive protein; NT-proBNP, N-terminal pro–brain natriuretic peptide.

Table 3 shows the volume and proportion of repeat CRP and NT-proBNP tests where the index test result was abnormal and the repeat was also abnormal. In 89% of previously abnormal repeat CRP tests and 97% of previously abnormal repeat NT-proBNP tests, no clinically informative change to the test result was observed.

TABLE 3.

Australian Medicare Benefits Schedule (MBS) Item-Based Fee of Each Test and Projection to Observed Volume

Test and repeat interval	Single test cost per MBS, AU$	Total No. of tests	Total No. of potentially inappropriate tests	Total cost for volume in the study period, AU$	Total cost for potentially inappropriate tests, AU$
Vitamin D	30.05	4495	413a	135,075	12,411
Lipid studies	11.65	2872	196a	33,459	2283
Iron studies (total cost including ferritin, serum iron and transferrin)	32.55	5475	369	178,211	12,011
Standalone ferritin	18.00	148	16	2664	288
CRP (within 3 days)	9.70	40,134	6947	389,300	67,386
NT-proBNP (within 7 days)	58.50	2780	179	162,630	10,472
Total		55,904	8120	901,339	104,851

CRP, C-reactive protein; NT-proBNP, N-terminal pro–brain natriuretic peptide.

^aConservative estimate.

The MBS cost associated with a single test and the total costs of unnecessary testing are listed in Table 4, with indicative total of costs linked with the number of tests.

When limiting the definition of inappropriate tests to repeat testing prior to recommended repeat testing intervals, the costs of inappropriate testing amounted to AU$105,000 as the conservative estimate (12% of all fees for the included tests). When including vitamin D testing and lipid testing, the total costs of inappropriate testing amounted to AU$259,000 as the liberal estimate (30% of all fees).

DISCUSSION

This study found potential overuse of vitamin D and lipid tests in the hospital setting, inappropriate selection of iron studies, and a large number of potentially inappropriate repeat testing of CRP and NT-proBNP. The target tests in our study were overused both due to likely unnecessary initial request, as evident in the case of vitamin D, lipids, and iron studies, and unnecessary repeat testing for all tests included. Most abnormal repeat CRP and NT-proBNP tests had an index test with an abnormal result, strongly suggesting that the repeat tests were unnecessary. The collective cost of all these unnecessary tests to the system under study ranged from AU$105,000 to AU$259,000. Although our study was limited to 2 hospitals, given there is AU$4 billion expended annually nationally under Australia’s Medicare system on pathology services, national costs from inappropriate use of pathology are likely considerable.³ Through our analysis, we have demonstrated the importance that pathology informatics plays in identifying pathology use and the quality of services provided to patients in hospital care.

Despite the existence of guidelines indicating that population-based vitamin D testing is not recommended in Australia,²⁹ as well as a systematic review that found there was little evidence supporting vitamin D screening and supplementation in acute care,³⁰ we found that ordering of vitamin D tests occurs in the hospital setting. Similar results have been reported elsewhere. In a Croatian study, it was reported that more than half of the vitamin D tests conducted in a tertiary setting were unnecessary.¹² Patients with malabsorption or osteoporosis may require testing, but generally, vitamin D supplements can be administered without the need for testing³¹; less than 1% of Australians have severe vitamin D deficiency.³² Although we defined a repeat vitamin D as inappropriate, it is likely that a large number of the initial vitamin D requests are also inappropriate in the hospital, and our findings indicate interventions to reduce unnecessary use are warranted.

While cardiovascular events are common in the hospital setting, lipid testing is not a recommended panel test in this setting.^33,34 Total cholesterol falls acutely and remains low for approximately 8 weeks following any acute medical illness,³⁵ which should be considered since a lipid test during this window may provide a misleading result and repeat tests will also not be beneficial. In acute coronary syndrome, an initial lipids profile is recommended within 24 hours or after at least 6 weeks have lapsed following such an incident as a prelude to the introduction of statin or other lipid-lowering therapies.³⁶

Depending on clinical indications, such as in diabetic ketoacidosis, hypertriglyceridemia, and acute pancreatitis, where elevated levels of triglycerides are expected, triglycerides are sometimes ordered as a standalone test instead of a lipids panel.¹⁷ We hypothesize that some of the repeated lipids testing we observed can be attributed to a panel-based test rather than selectively ordering a single triglyceride test.

Use of iron studies and ferritin tests is a potential area for intervention to reduce unnecessary tests, as we found that iron studies were ordered 36 times more often than standalone ferritin tests, even though ferritin levels are the best indicator of early iron deficiency.³⁷ Unbundling order sets into single tests has shown a reduction in inappropriate ordering of tests,³⁸ and we hypothesize test bundling may be a systematic influencer in the requesting behavior of iron studies as opposed to a standalone ferritin. In a study of inappropriate repeat tests in a teaching hospital in Turkey, 13% of ferritin tests and 16% of iron tests were inappropriately repeated within a 3-month period,³⁹ similar to our results. We found no other studies that examined the extent of overuse of iron studies compared to standalone ferritin tests. Further study of patient conditions may assist in explaining the variance to ordering behavior.

C-reactive protein is regularly ordered in acute care settings, often multiple times.⁴⁰ In a study in the Netherlands, it was reported that among patients who had a CRP test in the ED, 59% received a repeat test within 24 hours, and in 88% of these cases where repeats were ordered, there was no recorded change to patient management.⁴¹ Most of the repeat CRP tests in our study followed a previous test with an abnormal result and thus were likely redundant. Of the repeats that were conducted within 3 days of the previous ones, 83% were already abnormal before. While monitoring of treatment response may be driving the testing, the kinetics of CRP mean that a repeat test, where the first test was already abnormal, is unlikely to be substantially different.²²

The National Institute for Health and Care Excellence guidelines indicate the use of NT-proBNP testing when heart failure is suspected in a clinical episode.⁴² On the basis of these guidelines, the Royal College of Pathologists does not recommend a repeat test during a clinical episode.¹¹ With this in mind, the volume of repeat NT-proBNP tests, as well as repeat tests where the index test was abnormal, represents inappropriate pathology testing.

The focus on improving the use and quality of pathology services is an area of growing interest in Australia.⁸ Since 1999, the Australian government has established a Quality Use of Pathology Program (QUPP), which primarily focuses on evaluating and monitoring pathology usage across the country.⁴³ Some of the key initiatives funded under this program have reported on enhancing pathology use in general practice, the use of machine learning to improve pathology use, and reviewing standardized units and terminology.⁴³ Our study sits well within the aims of the QUPP.

For public patients in a public hospital in Australia, tests are funded by the state government, which is composed of taxpayer contributions.⁴⁴ This model of service may affect ordering behavior, as hospitals are reimbursed by the state for the costs associated with the test. Furthermore, as this is a national policy, our results may be generalizable to the hospitals across different states and regions, although practice may vary across settings, and this requires further investigation.

The role that the Royal College of Pathologists of Australasia (RCPA) plays in regulating pathology use is vital to improving and managing pathology resources and utilization.⁴⁵ The RCPA is responsible for the training and support of pathologists to improve the use of pathology for patients.⁴⁵ From a quality improvement lens, some of the strategies to better the state of pathology include collaborating with government entities and driving projects to standardize pathology reporting practices.⁴³ Furthermore, the formation of the Quality Use of Pathology Committee, which informs the QUPP and is a collaborative venture across different health-focused organizations, is promising.⁴⁶ This study has established a localized focus on potentially inappropriate pathology testing for the selected South Australian hospitals.

Our study had several limitations. The lack of data on date of admission and discharge led to the study being reliant on a definition of a single episode of care of 30 days from the index test. As a result, the rate of repeat tests in practice may be higher than determined in our study. We were unable to analyze patient clinical indications, which limits our ability to conclusively report a given test as “inappropriate,” as in some patients repeat tests may be appropriate.

Many issues are likely to contribute to the inappropriate use of pathology services, including clinical guidelines on test ordering⁴⁷; availability of decision support; physician knowledge, preferences, and expectations; patient morbidity; and health system structures. In this study, while limited to only 6 tests, the findings highlight the magnitude of potentially inappropriate pathology use in the hospital setting and the need for quality use of pathology programs to improve the value of testing in patient care. We also identify a need to explore the effectiveness of initiatives such as the Choosing Wisely campaign, guided by the RCPA, that recommends reduced testing of vitamin D, for example, and how well this is adhered to in different medical settings.⁴⁸ There is significant potential to inform best practice in pathology through data-led pathology informatics studies to drive positive patient outcomes.

Funding: This work was supported by The Hospital Research Foundation Group (THRF).

Conflict of interest disclosure: The authors have nothing to disclose.