Providers expect that patients with chronic conditions will complete necessary laboratory tests; 1 in 7 patients with diabetes did not do so within 6 months.
Published Online: May 21, 2011
Howard H. Moffet, MPH; Melissa M. Parker, MS; Urmimala Sarkar, MD, MPH; Dean Schillinger, MD; Alicia Fernandez, MD; Nancy E. Adler, PhD; Alyce S. Adams, PhD; and Andrew J. Karter, PhD
Objectives: To estimate rates and predictors of clinical laboratory test completion by patients with diabetes after provider referral.
Study Design: Prospective cohort study.
Methods: Among 186,306 adult members with diabetes in Kaiser Permanente Northern California, we searched the electronic medical records (July 1, 2008, to June 30, 2009) of each patient for the first outpatient order to obtain the following laboratory tests commonly used to measure risk factor control or adverse effects of pharmacotherapy: levels of glycosylated hemoglobin, low density lipoprotein cholesterol, serum creatinine, urinary albumin, or creatine kinase (the latter only among persons using statins). We measured laboratory test attendance as completion of an order (including time to results) within 6 months of the referral date and looked for variations by subgroups.
Results: Laboratory test attendance ranged from 86% for glycosylated hemoglobin level to 73% for serum creatinine level. Time to laboratory test attendance was a median of 7 to 11 days and a mean of 25 to 30 days. Laboratory test attendance was more likely for women and older patients or for orders after a face-to-face provider visit and was less likely for orders by a pharmacist. However, most variations (even by laboratory copayment) were small or not clinically substantive. In subanalyses, we observed no clinically significant variations by race/ethnicity, socioeconomic status, trust in provider, or patient–provider communication and found no association with depressive symptoms, health literacy, or English fluency.
Conclusion: The fact that 1 in 7 patients did not complete laboratory tests within 6 months of the provider referral may help explain why healthcare services seem to fall short of optimal diabetes care.
(Am J Manag Care. 2011;17(5):339-344)
One in 7 patients in this diabetes registry cohort did not complete laboratory tests within 6 months after provider referral; this may help explain why healthcare services seem to fall short of optimal diabetes care.
The median time from referral to laboratory test attendance was at least 1 week, and the mean time was almost 1 month.
No substantive race/ethnicity or other social differences in laboratory test attendance were observed; this suggests that disparities in rates of testing that have been observed among population-based studies may be attributable to variations in referral rather than patient adherence.
Laboratory testing of blood and urine provides essential objective evidence for making clinical decisions in diabetes care management.1 Testing is important not only for routine monitoring but also for determining the necessity and consequences of treatment intensification, which may be needed to achieve clinical goals2,3 or detect adverse drug reactions to prevent clinical harm.4
Routine tests of glycosylated hemoglobin (A1C), low-density lipoprotein cholesterol (LDL-C), serum creatinine (SC), and urinary albumin (UA) levels are recommended for= patients with diabetes.5 Rates of testing in diabetes care are routinely reported in journals6 or published by health plans as quality measures (eg, Healthcare Effectiveness Data and Information Set), with variations and disparities noted. For example, Schmittdiel et al7 found that measurement of A1C and LDL-C levels within 6 months after treatment intensification occurred less than half the time. Racial/ethnic disparities in A1C testing have also been reported in population-based studies,8,9 although differences are greatly attenuated or absent in managed care settings with fully insured patients.10 Nonadherence to laboratory test requests could contribute to higher rates of poor diabetes control among vulnerable patient populations. However, it is unknown whether disparities in rates of testing are attributable to variations in ordering by providers or to differences in laboratory test attendance by patients after a test is ordered.
This article examines adherence to clinical laboratory test requests. We calculated rates of laboratory test attendance by patients with diabetes after provider referral in an electronic medical record (EMR) system. We hypothesized that a small segment of the patient population fails to fulfill laboratory test orders and that nonfulfillment would not vary by type of test. We further hypothesized that vulnerable subgroups (eg, racial/ethnic minority groups or those with less education, lower income, limited English fluency, or inadequate health literacy) would be at highest risk of nonutilization because of latent barriers (eg, lack of transportation or inadequate understanding of the clinical importance of laboratory tests), as well as subgroups with low trust in their provider or poor patient–provider communication.
This study was conducted among patients in the diabetes registry at Kaiser Permanente Northern California (hereafter Kaiser). We identified a cohort of 186,306 adult patients with diabetes having continuous health plan membership from July 1, 2008, to December 31, 2009. At Kaiser, electronic order entry for laboratory tests (a component of the EMR) facilitates analysis of patient adherence to laboratory test orders by tracking separately (1) the provider’s order to the laboratory for each test and (2) the patient’s adherence to the provider’s request to attend the laboratory and provide specimens. Kaiser is a closed system, laboratory test orders are not transferred outside of Kaiser, and paper laboratory test slips are no longer used, so we were able to accurately capture the outcome for 100% of the cohort. Patients merely go to any Kaiser laboratory, where the laboratory test order awaits, and the ordered test will be conducted on the specimens provided.
In this cohort, we searched the EMRs of each patient for the first occurrence between July 1, 2008, and June 30, 2009, of an outpatient laboratory test order for the following tests of interest: levels of A1C, LDL-C, SC, or UA. To contrast routine diabetes tests with a test that might be more urgent, we also tracked creatine kinase (CK) testing among patients using statins (dispensed within the previous 6 months); this test is often ordered to detect an asymptomatic adverse effect of a medication or, more commonly in this clinical environment, to diagnose a possible adverse reaction among symptomatic patients.
We used individual laboratory test orders and laboratory panel orders (eg, diabetes panel and lipid panel) to capture orders for each test. We excluded standing orders (estimated to be about 5% of the A1C test cohort), which cannot be easily ascertained because they do not appear as new orders in the system; we excluded any order with an expected completion date (estimated to be 9% of the A1C test cohort), for which the provider has entered a future date for order completion and presumably has requested the patient to attend the laboratory around that future date (ie, not immediately). Therefore, all included orders were eligible for immediate fulfillment. Each patient could be counted in more than 1 laboratory test cohort.
Our outcome of interest was patient laboratory test attendance at a Kaiser laboratory within 6 months of each identified test order. To measure laboratory test attendance, we identified the first test result after a given order and used the date that the specimen (blood or urine) was collected as the attendance date. We censored follow-up at 6 months after the order date, although some orders may be valid for up to 1 year; because we do not know when the provider intended that the patient would attend the laboratory, we assumed that 6 months was a generous window for completion. Therefore, laboratory test attendance was affirmative if a result was entered within 6 months. We calculated laboratory test attendance as a percentage of laboratory test orders that were completed and counted the number of days between order and attendance.
We evaluated differences by sex, age, type of provider and visit (face to face vs other), and laboratory copayment. In subanalyses of patients who had participated in the 2005-2006 Diabetes Study of Northern California (DISTANCE) survey (n = 20,188),11 we further evaluated differences by race/ethnicity, annual income, educational attainment, employment status, depressive symptoms,12 health literacy,13 English fluency, trust in provider,14 and patient–provider communication (a patient reports that the provider explains things in a way the patient can understand).15 This study was approved by the Institutional Review Board of the Kaiser Foundation Research Institute.
Laboratory test attendance ranged from a high of 86% for A1C (138,017 tests) to a low of 73% for SC (139,962 tests) (Table 1). Time to laboratory test attendance was a median of 7 to 11 days and a mean of 25 to 30 days. Laboratory test attendance for CK, which we assumed might have some urgency, had a pattern identical to that for A1C. Survival curves of open laboratory test orders for each test were largely indistinguishable (data not shown).
In unadjusted analyses, women were slightly more likely than men to attend laboratory testing, but the male–female differences were minimal (<2% for any laboratory test) (P<.05). Older patients (>65 years) were more likely to attend laboratory testing than younger patients (age range, 19-34 years) (P <.001); the rate difference was 12% to 17%. Orders resulting from face-to-face visits were more likely to result in laboratory test attendance (P <.05), although differences were small (6%-13%). Orders entered by a pharmacist were less likely to be completed (1%-7% lower). There was no evidence that laboratory copayments were a significant barrier to laboratory test attendance.
In subanalyses among patients who had completed the DISTANCE survey, no laboratory test completion rates differed substantively across any of the factors we considered (<7% difference for any test), although some differences reached statistical significance (eg, variations in LDL-C testing rates were statistically significant, with the lowest rates among persons of white race/ethnicity) (Table 2). Some social patterns were not as predicted (eg, greater educational attainment, higher annual income, and employment had some association with lower laboratory test attendance), but no differences were large enough to be considered clinically substantive. We observed no statistically significant differences in laboratory test attendance among patients by depressive symptoms, health literacy, or English fluency.
To our knowledge, this is the first large study to use EMR-based laboratory test orders to estimate rates of laboratory test attendance by patients who are given a provider referral. A significant proportion of patients (14%-27%) did not attend the laboratory within 6 months, and the time to laboratory test attendance varied little by test, even when comparing routine A1C tests with potentially urgent CK tests. We observed no substantive social disparities in laboratory test attendance. Although we cannot generalize these results that were observed in a managed care setting to other populations, our findings suggest that any disparities in rates of clinical testing may be attributable to variations in referrals rather than patient adherence. Much has been made of the concept of clinical inertia, the finding that medication changes are not made in a prompt manner, but delays in attending laboratory testing represent another barrier to timely care. Clinical action cannot be taken on laboratory tests that have not been completed.
PDF is available on the last page.