| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Evidence-based Medicine and Test Utilization |
1 Clinical Epidemiology Program, Ottawa Health Research Institute, C405, Ottawa Hospital, Civic Campus, 1053 Carling Ave., Ottawa, ON K1Y 4E9, Canada.
2 Department of Pathology, Queen’s University, Kingston, Ontario, Canada.
aAuthor for correspondence. Fax 613-761-5492; e-mail carlv{at}ohri.ca.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: We performed a cross-sectional study using high-quality, population-based clinical databases that included adults in Eastern Ontario, Canada, between September 1999 and September 2000 for incidence of repeating eight common laboratory tests (hemoglobin, sodium, creatinine, thyrotropin, total cholesterol, HDL-cholesterol, ferritin, and hemoglobin A1C). Tests were classified as potentially redundant if repeated within the test’s baseline testing interval. For creatinine, sodium, and hemoglobin, only tests repeated in the community were considered. For a sensitivity analysis, we varied the repeat interval by 25%, excluded tests repeated by different physicians, and excluded repeats of normal tests.
Results: Almost 4 million tests were conducted during the study year. Most tests (76%) were conducted on patients in the community. More than one-half of all people in the population had at least one laboratory test, with an overall testing rate of 367 tests per 100 people per year. Repeat testing within 1 month accounted for 30% of all utilization (109 repeat tests per 100 people per year). Repetition was more common in hospitalized patients, varied extensively among tests, and was concentrated in a limited number of people. For the eight tests included in the study, charges of potentially redundant repetition in adults totaled between $13.9 and $35.9 million (Canadian) annually.
Conclusions: Laboratory test repetition is very common, makes up a significant component of overall test utilization, and is costly.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Previous investigations have studied repeat testing in the community (21), in hospitals (17)(22)(23)(24), or when patients are transferred from the community into the hospital (25)(26)(27). Over an 8-month period in The Netherlands, 38% of common laboratory tests in the community had been conducted previously during that time period (21). Test repetition is also common if patients are admitted to hospital. Rix and Stump (26) found that only 3.1% of tests conducted during the first day of hospitalization had been performed in the week before admission; however, 71% (25) and 63% (27) of tests conducted in the week and year before admission, respectively, were repeated during the hospitalization. Repeat testing is very common during hospitalization. Dixon and Laszlo (23) found that the tests on almost all inpatients were repeated, even when the initial chemical profile was normal. At the Brigham and Women’s Hospital, 28.2% of 78 798 common laboratory tests were repeated unnecessarily early (24). Valenstein et al. (22) found that 27% of 936 inpatient, non-intensive care unit chemistry profiles had been measured previously that day (17). In the intensive care unit, 82.8% of electrolyte profiles of 145 patients had been measured previously that day (22).
To accurately measure repeat laboratory testing, a population-based assessment is optimal. This allows laboratory utilization to be studied for everyone within a geographic area, rather than within a particular hospital or health services organization. A population-based analysis allows laboratory utilization to be followed even when patients transfer between different sectors of the healthcare system, such as from the community to the hospital. Finally, a population-based analysis produces unbiased utilization rates because a true denominator (i.e., all people in a particular area) rather than a "pseudo denominator" (i.e., all people who had a laboratory test) is used. This is necessary to meaningfully compare repeat laboratory testing to the utilization of other health services. This study is a population-based examination of repeat laboratory testing in the community and hospital for people in Eastern Ontario, Canada.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
laboratory tests
In Ontario, medical laboratory testing is provided by private and hospital-based laboratories. With some exceptions, the former provide laboratory testing exclusively to community-based patients. Hospital-based laboratories provide testing primarily to hospitalized patients but also to community patients who attend hospital-based clinics or in areas where private laboratories are unavailable.
All private laboratories operating in Eastern Ontario participated in the study (see Appendix A). All hospitals in the area participated as well (Appendix A) except for those in Hawkesbury and Cornwall. This was primarily attributable to technical limitations of the laboratory information systems within these institutions that precluded them from participating in the study. On the basis of the data collected for the study, we estimate that our sample missed 1.5% of all tests in Eastern Ontario. The study was approved by the ethics review board of the Ottawa Hospital as well as the ethics review boards or administrative representatives of all participating institutions.
Because a study of all tests would have been infeasible, we examined the utilization of eight laboratory tests. These included a mixture of common (including hemoglobin, creatinine, and sodium) and less common (thyrotropin, total cholesterol, HDL-cholesterol, ferritin, and hemoglobin A1C) laboratory tests. We believed that including such a mixture would increase how representative our sample was for the entire population.
Between September 1, 1999, and September 1, 2000, participating laboratories downloaded the patient identifier, test date, test result, and ordering physician of all study tests from their laboratory information systems. Patients were identified by their Ontario Health Insurance Plan number. Tests were collated into computer files that were encrypted and transferred electronically to the Institute for Clinical Evaluative Sciences. Here a person with written permission to see unencrypted patient identifiers decrypted the entire data file. The Ontario Health Insurance Plan numbers and physician identifiers were then individually encrypted, and the file was stored on a stand-alone and secure computer network. Using these procedures, researchers could study the data file without ever seeing unencrypted patient or physician identifiers. Because it used the same encryption key for patient and physician identifiers, the data file could also be linked to other administrative databases.
database linkage
The Registered Patient’s Database was used to identify each person’s age and sex. The Registered Patient’s Database records basic demographic information for each Ontarian. To determine whether a test was conducted while the patient was in the hospital, we linked to the Discharge Abstract Database, which records the admission and discharge date of all hospitalizations (including same-day surgeries) in Ontario. Tests conducted between the admission and discharge dates were classified as hospital tests. To determine whether a test was conducted while the patient was in the emergency department, we linked to the Physician Services Database, which records claims for physician services. Services conducted in emergency departments are identified by special claim codes. Tests that were ordered on the same date by the same physician who claimed an emergency assessment were classified as emergency tests.
analysis
Using each laboratory’s reference intervals, we determined whether test results were abnormal. Because each laboratory’s testing methodology and reference intervals for each test were similar, we concluded that it was unnecessary to standardize test results. Tests were excluded if there was a missing or invalid patient identifier, if the test had a missing or invalid result, or if patient sex or age was missing. The latter was needed to determine whether the test had a normal result. Because we concentrated on laboratory utilization in adults, tests on patients <18 years of age were also excluded.
To estimate the population of the study, we used 1996 Canadian Census data categorized by the Forward Sortation Area (FSA). The borders of three rural area FSAs (K0K, K0H, and K0A) were almost identical to the border for our study area. We summed the population of people 18 years and older of these FSAs and all other Ontario FSAs to the East. We adjusted the 1996 population estimate to 2000 values by use of multipliers that were derived from changes in county-level, age-cohort-specific population changes between 1996 and 2000. The Kaplan-Maier method was used to measure test repetition rates (28). SAS, Ver. 8.2, was used for all analyses.
estimating province-wide charges for potentially redundant laboratory tests
We estimated the province-wide charges for potentially redundant laboratory testing. We first used census estimates of the adult population of the study area (1.09 million) and that of the entire province (9.16 million) to estimate the total number of study tests for the entire province. We used a third-party payer perspective to calculate charges for potentially redundant laboratory testing. In Ontario, payment for laboratory tests is determined by the Ministry of Health (29). For this study, charges for hospital tests were identical to those for private laboratories, although such tests are paid out of the hospital’s global budget and not by the Ministry.
Repeat tests are often necessary for proper patient care (30). Determining whether particular tests are appropriate can be difficult, even when primary data are available (6). To estimate the proportion of tests that are potentially redundant, we used the repeat intervals for each presented in Appendix B to determine the time within which test repetition was potentially redundant. Whenever tests could be used for different indications, we always chose the shorter repeat interval to define potentially redundant. As a sensitivity analysis, we varied these repeat intervals by a relative increase and decrease of 25%. We used Kaplan-Maier survival curves (28) to determine the proportion of tests repeated within each repeat interval, which was multiplied with the provincial total number of tests to calculate charges. The Kaplan-Maier curves were created for each test in four conditions: (a) including all tests repeated within the repeat interval; (b) excluding tests repeated by the same physician (because physicians should consider the results of each test they order and make appropriate decisions, including whether the test should be repeated, based on this information); (c) excluding repeats where the index was abnormal [because it is often good medical practice to repeat abnormal tests (31)]; and (d) excluding tests that meet either of these criteria. Finally, patients who are seen in the emergency department or in hospital are frequently acutely ill; in these cases, assessment and monitoring of important laboratory values such as sodium, creatinine, and hemoglobin is appropriate. We therefore did not consider such tests as potentially redundant if ordered while the patient was in the emergency department or the hospital.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
). Hemoglobin, creatinine, and sodium were the most common, accounting for 68.1% of all tests. These tests were also more likely to be conducted in the emergency department or in the hospital. Cholesterol and hemoglobin A1C were least likely to have normal results (45.9% and 31.9%, respectively).
|
Population-based test utilization rates and the proportion of the population having at least one test during the study year are presented in Table 2
. The eight study tests were commonly used with an overall rate of 366 tests per 100 people per year. Test rates ranged from 9 per 100 people-years for ferritin to 108 tests per 100 people-years for hemoglobin. Test rates varied extensively among age groups, with test utilization being highest in the eldest quartile for all tests except total and HDL-cholesterol. Slightly more than one-half of the population had at least one test during the study year. This proportion varied extensively among tests (6% for ferritin to 45% for hemoglobin). The proportion of adults having one or more tests increased with age, with almost three-fourths of adults over 65 years of age having at least one of the study tests.
|
repeat testing
The probabilities that tests were repeated during the study year are shown in Fig. 1A
. Overall, tests were repeated very rapidly at the start of observation, with 22% of tests being repeated within 1 week. By the end of the year, 63% of all tests had been repeated. Test repetition varied extensively among the tests. Common tests, such as hemoglobin and sodium, were repeated very quickly at the start of observation. Other tests, most notably that for hemoglobin A1C, had a repetition curve that was sinusoidal in shape. At the end of the test year, the probability that tests were repeated ranged from 40% for ferritin, thyrotropin, and total and HDL-cholesterol to 74% for sodium.
|
Two factors affecting test repetition are illustrated in Fig. 1B
. The location and the result of the index test strongly influenced time to test repetition. Index tests conducted while the patient was in the hospital or the emergency room were repeated much more quickly than when patients were in the community. In each location, tests were repeated more quickly when the index test was abnormal.
The population-based measures of repeat testing are presented in Table 3
. When all tests were considered together, we found a 1-week test repetition rate of 80 tests per 100 people per year. This accounted for 21.7% of all test utilization. During the year, 23% of the population had a repeat test. As the repeat interval increased, the test repetition rate also increased, with a 1-year repetition rate of 229 tests per 100 people per year. Repetition rates varied extensively among tests. For example, at 1 week, tests with the overall highest utilization rates, including hemoglobin, creatinine, and sodium, had repeat testing rates that were considerably greater than the rates for the other tests (Table 3
).
|
potentially redundant repetition
The total annual charges for potentially redundant utilization of the eight study tests for adults in the province are presented in Fig. 2
. Extrapolating our data to the entire province, we estimated that there would be a total of 33.5 million study tests annually (365 tests per 100 persons per year) with total charges of $215.2 million (Canadian).
|
With the baseline testing repeat presented in Appendix B, potentially redundant rates varied between 49.4 and 153.5 tests per 100 persons per year, depending on the exclusion criteria and the repeat testing interval used to label tests as potentially redundant. With no exclusions, redundant testing rates varied from 4.4 tests per 100 for HDL to 33.5 tests per 100 for creatinine.
For the tests included in this study, potentially redundant laboratory repetition consumed between $13.9 million and $35.9 million annually, depending on the exclusion criteria and the repeat interval that was used (Fig. 2
). This comprises 6.5–16.7% of total charges for the study tests annually.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Laboratory utilization consumes considerable resources, even when direct costs alone are considered (1)(2)(3)(4)(5). Repeat tests are a special component of laboratory utilization because they could be particularly conducive to modification, especially by improving the dissemination of results among physicians (5)(6)(15)(16)(17)(18). Our study shows that tests are frequently repeated. Of the tests included in our study, repeats within 3 months accounted for 40.4% of overall test utilization. Even with conservative criteria, we found that potentially redundant tests are common. Despite the small unit cost of these tests, the frequent use of them brought the total annual charges of potentially redundant repetition up to $35.9 million. It should be noted that charges for medical technologies are generally much higher than costs in the United States, especially with regard to laboratory tests. The frequency and cost of repeat testing indicate that overall test utilization could be heavily influenced by targeting repeat testing alone.
To design and improve interventions that influence test repetition, we should identify the patient and system factors that are associated with repeat testing. In this report, we identified the location and normality of the index test as two factors that strongly influenced time to repetition. Other patient, physician, and test factors could influence repeat testing, such as patient age, severity of illness, whether the repeating physician ordered the initial test, location of patient assessment, and test type. Further research in this area could help explain why tests are repeated and help us to expand on interventions that have been shown to reduce the frequency or charges associated with repeat tests, as illustrated in two published studies (19)(32).
There are interesting comparisons between our study and previous investigations of laboratory repetition. One previous study of laboratory repetition in the community found a repetition rate of 38% over an 8-month period (21). We noted a considerably higher repetition rate of 56% as determined from the Kaplan-Maier curve in Fig. 1A
. This higher estimate is likely attributable to the inclusion of all healthcare sectors rather than just community patients. In addition, use of Kaplan-Maier methodology avoided underestimating repetition when index tests were conducted late in the study period. Our study confirms that test repetition is especially common for hospitalized patients (17)(23)(24), although some of this repetition is appropriate because of the acuity of hospitalized patients.
Potentially redundant laboratory tests are an especially important component of repeat testing because they might be especially conducive to interventions that decrease utilization. Although our study shows that potentially redundant tests are relatively common, there are several reasons that our study could underestimate the potential savings from abolishing such tests. The first reason is that our study only included eight common tests. If other laboratory tests have similar redundant rates, the potential savings from decreasing the utilization of all tests would be considerable. The second reason is that our criteria for "potentially redundant" were inexact. It is possible that tests that do not meet our criteria were also redundant. Although it is also very possible that tests deemed by our criteria as potentially redundant were completely valid, we believe that our estimates for redundant test repetition rates and charges are likely conservative given the conservative repeat intervals that we chose for the study (Appendix B). Despite these uncertainties, we believe that our study is still important for achieving an approximate measure of redundant testing. The third reason is that our study missed laboratory tests from three hospitals in the study area. Although tests in these centers likely comprised only 1.5% of all such tests, test repetition rates in our study will be underestimated without them.
Our study has other limitations. One is that, given the scope of the study and its reliance on administrative data, we had no clinical information surrounding each test. Such information could have made classification of test redundancy, which was admittedly crude in this study, more reliable. Such data, perhaps along with prospectively collected information, would also allow us to determine whether index tests were clinically indicated, further enhancing our classification of redundant testing. Another limitation is that we measured test charges and not costs, which can be divergent (33). Costs, along with the expense of resources consumed because of the repeat tests, would be required to accurately assess the consequences of redundant repeat testing. A third reason is that we extrapolated utilization rates from Eastern Ontario to the rest of the province to estimate total charges. Because utilization rates for other healthcare resources can vary among sectors of the province (34)(35), we are uncertain of the accuracy of our extrapolations to the entire province.
Despite these issues, our study shows that repeat laboratory tests make up a considerable proportion of overall test utilization. It also shows that potentially redundant laboratory utilization is common and costly. We believe that these observations show that repeat testing is an area of utilization that could be targeted with interventions to decrease repeat testing and decrease health expenditures.
|
|
|
Acknowledgments |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  C. van Walraven, N. Oake, P. S. Wells, and A. J. Forster Burden of Potentially Avoidable Anticoagulant-Associated Hemorrhagic and Thromobembolic Events in the Elderly Chest, May 1, 2007; 131(5): 1508 - 1515. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. van Walraven, G. Cernat, and P. C. Austin Effect of Provider Continuity on Test Repetition Clin. Chem., December 1, 2006; 52(12): 2219 - 2228. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Miyakis, G. Karamanof, M. Liontos, and T. D Mountokalakis Factors contributing to inappropriate ordering of tests in an academic medical department and the effect of an educational feedback strategy Postgrad. Med. J., December 1, 2006; 82(974): 823 - 829. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. Woodward, C. van Walraven, and J. E. Hux Utilization and outcomes of HbA1c testing: a population-based study Can. Med. Assoc. J., January 31, 2006; 174(3): 327 - 329. [Full Text] [PDF]
|
|
|
|
 |
|
 A. X. Garg, M. Mamdani, D. N. Juurlink, C. van Walraven, and for the Network of Eastern Ontario Medical Laborat Identifying Individuals with a Reduced GFR Using Ambulatory Laboratory Database Surveillance J. Am. Soc. Nephrol., May 1, 2005; 16(5): 1433 - 1439. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. R. Shah, J. E. Hux, A. Laupacis, B. Zinman, and C. van Walraven Clinical Inertia in Response to Inadequate Glycemic Control: Do specialists differ from primary care physicians? Diabetes Care, March 1, 2005; 28(3): 600 - 606. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
