When Do Clinicians Follow-up Abnormal Liver Tests in Primary Care?

Abstract

Background:

Many guidelines addressing the approach to abnormal liver chemistries, including bilirubin, transaminases, and alkaline phosphatase, recommend repeating the tests. However, when clinicians repeat testing is unknown.

Methods:

This retrospective study followed adult patients with abnormal liver chemistries in a patient-centered medical home (PCMH) from 2007 to 2016. All PCMH patients possessing at least one abnormal liver test (total bilirubin, aminotransferases, and alkaline phosphatase) were included. Patients were followed from the index abnormal liver chemistry until the next liver test result, or the end of the study period. The primary predictor variable of interest was the number of abnormal chemistries (out of 4) on index testing. Demographic and clinical variables served as other potential predictors of outcome. A Cox proportional hazards model was applied to investigate associations between the predictor variables and the time to repeat liver chemistry testing.

Results:

Of 9,545 patients with at least 2 PCMH visits and 1 liver test abnormality, 6,489 (68%) obtained repeat testing within 1 year, and 80% of patients had follow-up tests within 2 years. Patients with multiple abnormal liver tests and those with higher degrees of abnormality were associated with shorter time to repeat testing.

Conclusions:

A large proportion of patients with abnormal liver tests still lack repeat testing at 1 year. The number of liver tests abnormal and degree of elevation were inversely associated with the time to repeat testing.

Introduction

Liver chemistry tests are frequently ordered in clinical practice, and often yield abnormal results.^1–4 Liver test abnormalities are associated with an increased risk of liver disease, and given the insidious presentation of most early liver disease, follow-up of abnormalities is likely an indicator of high quality care, even in the absence of signs and symptoms.^5–8 Multiple guidelines and recommendations for addressing abnormalities exist in the literature, yet approaches utilized by physicians vary.^9–16 After gathering information regarding patient symptoms, signs, medication history, and risk factors for liver disease (i.e. alcohol abuse, drug use), the most common initial recommendation is to repeat liver chemistries to confirm or assess the trajectory of the abnormality.^10,11,17 However, timelines for repeat testing are absent, or lack consistency.

In this study, we addressed when clinicians repeat liver chemistries after an initial abnormal test. We have hypothesized that abnormal liver tests would be repeated within 1 year and that demographic, clinical, comorbidity, and liver test-specific variables would impact the time to follow-up testing. In particular, we hypothesized that clinicians would follow-up liver test panels with multiple concurrent abnormalities more quickly than panels with lone test elevations. Using administrative electronic health record (EHR) data from a single, primary care patient centered medical home (PCMH), we examined the practice of repeating liver tests in response to initial abnormalities. We analyzed factors associated with time to follow-up testing including the number of liver chemistry tests initially abnormal, as well as patient demographics, clinical variables, and comorbidities.

Methods

Study Design and Setting

This retrospective cohort study followed patients with abnormal liver chemistries who received care within an academic, internal medicine patient-centered medical home (PCMH) from January 1, 2007 to June 30, 2016. The internal medicine PCMH at the Medical University of South Carolina (MUSC) cared for over 30,000 unique patients during the study period with nearly 38,000 patient visits per year. Administrative data from the electronic health record (“EHR,” Epic© Systems Corporation, WI) and the clinical data warehouse at MUSC served as the primary data sources for this study. Administrative data included all inputs generated within the integrated MUSC system including the ambulatory, emergency room (ER), and inpatient settings. Data does not include testing or results obtained by external care providers and systems. Throughout the study period, demographic and clinical data were stored electronically in a physician-accessible EHR.

Study Sample

Patients included all adults (≥ 18 years of age) who visited the PCMH and had at least 1 abnormal liver test. Liver tests included: total bilirubin (Bill), aspartate aminotransferase (AST), alanine aminotransferase (ALT), and alkaline phosphatase (ALP). Liver tests were identified as abnormal if any one of these aforementioned tests exceeded the upper limit of normal (ULN) for the reference range in the lab at MUSC: total bilirubin > 1.2 mg/dL, AST > 34 IU/L, ALT > 45 IU/L, and ALP > 150 IU/L. Further eligibility criteria for inclusion was the requirement that patients must have visited the PCMH on at least 2 occasions, in order to account for those patients referred to the tertiary care medical center, but electing to resume care elsewhere. Patients meeting these criteria entered the cohort at the time of their earliest abnormal liver test, and were followed longitudinally throughout the study period.

Study Outcomes

Patients were followed from the date of the index abnormal liver test (occurring between 2007 and 2016) until follow-up testing, death, or termination of the period of observation (June 30, 2016). Follow-up testing was defined as the next set of values (results, both normal and abnormal) for liver chemistries identified within the dataset following the initial abnormality. Liver chemistries ordered both in bundled order sets (i.e. panels including bili, AST, ALT, ALP) and by individual test entity fulfilled the outcome measurement. Test orders without results (situations in which patients did not obtain testing) were not included in the outcome assessment.

Predictor Variables

Demographic, clinical, and comorbidity variables served as potential predictors of outcomes (Table 1). Age was divided into 3 categories: 18 to 44, 45 to 64, and 65 or older. Categories of smoking status included those patients currently smoking, previously quit smoking, and either never smoked or did not receive documentation. The ordering provider of the originally abnormal liver tests was dichotomized into those practicing in the clinic (PCMH provider at the time of the abnormality) and those not.

Table 1.

Cohort Demographics by the Number of Abnormal Liver Tests on Index Test

	1 Abnormal Test (n=6,155)	2 Abnormal Tests (n=2,684)	3 Abnormal Tests (n=562)	4 Abnormal Tests (n=144)	p-value
Age (years; mean, SD)	54.7 ± 16.6	52.5 ± 15.8	52.5 ± 16.9	53.7 ± 19.0	<0.001
Male (%)	2726 (44.3%)	1272 (47.4%)	286 (50.9%)	57 (39.6%)	0.0011
Race (%)					0.003
Black	2789 (45.3%)	1223 (45.6%)	300 (53.4%)	75 (52.1%)
White	3236 (52.9%)	1403 (52.3%)	245 (43.6%)	65 (45.1%)
Other	130 (2.1%)	58 (2.2%)	17 (3.0%)	4 (2.8%)
Unmarried (%)	3327 (54.1%)	1482 (55.2%)	336 (59.8%)	144 (60.4%)	0.028
Public Ins. (%)	3497 (56.8%)	1445 (53.8%)	336 (59.8%)	79 (54.9%)	0.018
Distance (miles; mean, SD)	31.0 ± 149	27.2 ± 101	37.0 ± 148	26.6 ± 65	0.001
Far to MUSC >50miles	572 (9.3%)	261 (9.7%)	336 (13.0%)	19 (13.2%)	0.018
Poverty (%)	1993 (32.4%)	857 (31.9%)	203 (36.1%)	53 (36.8%)	0.172
PCMH Provider (%)	1009 (16.4%)	382 (14.2%)	82 (14.6%)	21 (14.6%)	0.064
Smoking Status (%)					0.044
Never	5288 (85.9%)	2295 (85.5%)	479 (85.2%)	118 (81.9%)
Quit	590 (9.6%)	229 (8.5%)	55 (9.8%)	15 (10.4%)
Current	277 (4.5%)	160 (6.0%)	28 (5.0%)	11 (7.6%)
Obesity (%)	2134 (34.7%)	1009 (37.6%)	188 (33.5%)	34 (23.6%)	0.001
Hypertension (%)	2158 (35.1%)	897 (33.4%)	231 (41.1%)	61 (42.4%)	0.001
Diabetes (%)	966 (15.7%)	424 (15.8%)	121 (21.5%)	35 (24.3%)	<0.001
Cancer (%)	466 (7.8%)	195 (7.3%)	188 (10.1%)	25 (17.4%)	<0.001
Myocardial	247 (4.0%)	130 (4.8%)	31 (5.5%)	7 (4.9%)	0.159
Infarction (%)
Liver Disease (%)	130 (2.1%)	119 (4.4%)	77 (13.7%)	29 (20.1%)	<0.001
Mental Health (%)	984 (16.0%)	444 (16.5%)	103 (18.3%)	21 (14.6%)	0.465
Alcohol Abuse (%)	267 (4.3%)	157 (5.9%)	61 (10.9%)	17 (11.8%)	<0.001
Hyperlipidemia (%)	1428 (23.2%)	582 (21.7%)	136 (24.2%)	35 (24.3%)	0.354

Index abnormality out of 4 tests, including total bilirubin, AST, ALT, and alkaline phosphatase;

^*Smoking history includes current and past smokers

^**Public insurance includes Medicare plus Medicaid

^***Poverty is determined by zip code of the patient’s residence was matched to 2010 Census data

^†Mental health includes a composite of depression and anxiety

^‡Alcohol abuse based upon ICD-9/10 code; Data for comorbidities come from Internal Classification of Diseases, Ninth and Tenth Revisions, Clinical Modification (ICD-9/10-CM) and derived by a modified Elixhauser coding algorithm.

The abnormal liver tests for each patient were categorized by the number of tests exceeding the reference range on index testing (out of 4 tests: Bili, AST, ALT, and ALP). Test abnormalities were further classified based upon the degree of elevation relative to the upper limit of the normal reference range (as noted previously): 1–2 times the upper limit of normal (1-2 × ULN), 2-4 × ULN, and > 4 × ULN.

Accurate data regarding social determinants of care were limited, so the zip code of the patient’s residence was matched to 2010 Census data as a proxy for poverty status. The poverty variable is dichotomous and given a value of 1 if the zip code has ≥ 25% of its residents below the federal poverty level. Transportation barriers were modeled by determining distance to clinic (a continuous variable), which reflects the distance in miles between the center of the patient’s zip code and the medical campus.

Comorbidities selected for analysis were chosen based upon pre-existing liver disease, inherent risk factors of liver disease (such as metabolic syndrome for non-alcoholic fatty liver, and alcohol use), diagnoses with high rates of medication monitoring (mental health and statin use in coronary artery disease), and systemic disease that may impact the liver (cancer). Definitions and data for comorbidities came from Internal Classification of Diseases, Ninth and Tenth Revisions, Clinical Modification (ICD-9/10-CM) and were derived by a modified Elixhauser coding algorithm.¹⁸ Diagnoses included in the Elixhauser definition of “liver disease” include chronic viral hepatitis, non-alcoholic fatty liver disease (NAFLD), esophageal varices, liver necrosis, chronic liver disease and cirrhosis, complications of cirrhosis, unspecified liver disorders, and liver transplant status. Comorbidities were coded in dichotomous fashion. Obesity was coded dichotomously with a threshold of BMI ≥ 30 kg/m² designated as obese.

Analysis

Patient demographic, clinical, and comorbidity variables were compared by the number of abnormal liver tests on index testing. Direct comparisons of patient characteristics were made with chi-square tests for categorical variables and t-tests or Mann-Whitney U tests for continuous variables (2-sided P-values reported).

A Cox proportional hazards model was applied to investigate associations between the predictor variables and the time to repeating liver chemistries. The model outcome focused on the days between the initial liver test abnormality and results of the next test. P values < 0.05 were deemed significant. We computed hazard ratios and 95% confidence intervals in SAS 9.3 (SAS, Cary, NC) using PHREG procedure.

Results

A total of 9,545 unique patients with at least 1 abnormal liver test and at least 2 PCMH visits met inclusion criteria. The majority of patients (n = 6155, 64.5%) had only 1 abnormal liver chemistry on index testing. The mean age of study subjects was 54 years, and other demographic variable information is detailed in Table 1. Notably, there were statistically significant differences in the number of abnormal liver tests on index testing for patients with co-morbidities including hypertension, diabetes mellitus, cancer, liver disease, and alcohol abuse. Patients with these conditions tended to have more abnormal tests than those patients without, particularly for disorders associated with the metabolic syndrome such as diabetes mellitus and hypertension.

Of patients meeting entry criteria, 6489 (68%) had a repeat set of liver chemistries within 1 year of the initial abnormality. The proportion of patients with repeat testing increased over time (that is, with follow-up performed) was 38% at 3 months, 53% at 6 months, and 80% at 2 years. We also found that 1119 patients (11.7%) did not undergo repeat liver chemistry testing at any point during the study period.

The number of liver tests abnormal on index testing was significantly associated with follow-up. That is to say that the more tests that were abnormal, the more likely repeat liver tests would be performed. Time to event curves were significantly different for patients with 1, 2, 3, and 4 index liver test abnormalities, respectively and sequentially (Figure 1). Additionally, the greater the degrees of abnormality, the sooner repeat liver tests were obtained. For each abnormal liver test, time to event curves differed by the degree of initial abnormality, with higher values relative to the reference range followed up more rapidly (Figure 2). Male gender, poverty status (by zip code proxy), initial PCMH ordering provider, and previous smoking history were associated with a reduced likelihood in patients obtaining repeat testing over time (determined by Cox proportional hazard modeling for time to repeat liver testing (Table 2)). The presence of obesity, diabetes, cancer, liver disease, and depression were all associated with a higher likelihood of follow-up to abnormal liver tests.

Figure 1:

Time to repeat testing by the # of abnormal liver tests. Kaplan-Meier time to repeat test curves for the number of abnormal liver tests (out of 4) identified on index testing. Abbreviations: Total_ABN = total tests abnormal (n=1, 2, 3, or 4). Days = days to 1^st follow-up test.

Figure 2:

Time to repeat testing by the degree of abnormality for each liver test. Kaplan-Meier time to repeat test curves for the degree of abnormality of each liver test (out of 4) identified on index testing.

Table 2.

Cox Hazard Model

	Estimate	Hazard Ratio (HR)	Lower HR	Upper HR	P Value
Age	0.005	1.005	1.004	1.007	<.0001
Nonwhite	0.082	1.086	1.036	1.138	0.0006
Male	−0.082	0.921	0.881	0.963	0.0003
Unmarried	0.061	1.063	1.015	1.113	0.0101
Poverty	−0.040	0.961	0.915	1.009	0.1068
Distance to MUSC>50mile	0.019	1.020	0.946	1.099	0.6101
PCMH Provider	−0.410	0.663	0.622	0.708	<.0001
Total N of ABN_LFT (Ref. N=1)
N=2	0.182	1.199	1.142	1.259	<.0001
N=3	0.357	1.429	1.304	1.568	<.0001
N=4	0.572	1.772	1.489	2.109	<.0001
Smoking Status (Ref. Never)
Quit	−0.119	0.888	0.816	0.966	0.0054
Current	−0.231	0.794	0.711	0.885	<.0001
Obesity	0.058	1.059	1.011	1.110	0.0158
Hypertension	−0.016	0.984	0.925	1.048	0.6197
Diabetes	0.088	1.093	1.019	1.172	0.0130
Cancer	0.521	1.684	1.548	1.831	<.0001
Myocardial Infarction	0.100	1.105	0.988	1.236	0.0798
Liver Disease	0.504	1.655	1.475	1.858	<.0001
Mental Health	0.113	1.119	1.049	1.195	0.0007
Alcohol Abuse	0.080	1.083	0.978	1.199	0.1251
Hyperlipidemia	0.057	1.059	0.990	1.133	0.0976

Table 3 details the results of follow-up liver tests. Nearly half (48.6%, 4638/9545) of patients had no liver test abnormalities on repeat testing, nearly a third (32.7%) had the same number or more abnormalities, and 11.7% were without repeat testing.

Table 3:

Number of Abnormal Liver Chemistries on Repeat Testing by the Number of Abnormal Chemistries on Index Testing

		Number of Abnormal Liver Tests at Index
		1	2	3	4	Total
	0	3377	1099	148	14	4638
Number of Abnormal Liver Tests on Repeat	1	1425	415	64	12	1916
	2	498	726	123	10	1357
	3	73	124	166	29	392
	4	9	18	17	68	112
	Missing	773	302	44	11	1130
	Total	6155	2684	562	144	9545

Discussion

In this study of primary care patients and abnormal liver test follow-up, 68% of patients had repeat liver chemistry assessment within 1 year after the index abnormality. The greater the number of abnormal liver tests, and the more abnormal the liver test was, the more rapidly the repeat testing was performed (Figures 1 and 2). Multiple demographic, laboratory, and comorbidity variables had strong associations with the attainment and timing of follow-up testing. These data suggest that clinicians likely consider a combination of factors when ordering repeat liver chemistries; in addition, a number of these and other factors may influence the likelihood of patients obtaining follow-up testing.

Despite recommendations to repeat testing in response to liver chemistry abnormalities, little guidance as to the timing of reassessment is currently available. Liver chemistries themselves are highly reproducible, thus the utility of repeat testing lies not only in the confirmation of abnormality, but perhaps more importantly in the ability to assess the trajectory of the tests.¹⁹ As such, the time between tests holds clinical relevance, and this study identifies significant variability in time to repeat testing. Different contexts likely play some role in this variability and include the presence and severity of patient symptoms (or lack thereof) and the reasons for initial testing. However, with the penetration of EHR into clinical practice and the exponential growth in information delivery, physicians may receive abnormal results without the immediate availability of clinical context. Thus, understanding what factors (particularly those available to clinicians within the EHR) are associated with repeat liver chemistry testing may prove useful in developing models for high quality care delivery. This observational study provides insight into the current state of practice in primary care patients with abnormal liver tests.

Surprisingly, only 68% of patients had follow-up tests in the first year following abnormality, and only 80% had repeat liver chemistries within 2 years of the index testing. Potential reasons for these absent or delayed follow-up tests include the decision to pursue other avenues of investigation, such as robust history taking, focused laboratory testing (i.e. viral hepatitis serologies), or imaging. With these more direct evaluations for liver disease, clinicians may elect to forgo additional liver chemistry testing. Secondly, and perhaps as result of these other investigations, clinicians may make diagnostic conclusions without repeat testing, reflected by the relationship of alcohol use to repeat testing (HR = 1.07, 95% CI 0.97 - 1.19) in this study. If a diagnostic conclusion explains a liver test abnormality, repeat testing may not be necessary. For instance, most of (95.1%, 528) the 555 unique patients with lone elevations of bilirubin had values less than twice the upper limit of normal (< 2.4 mg /dL). Clinicians may have suspected a diagnosis of Gilbert’s syndrome, a benign genetic condition of disordered bilirubin metabolism affecting 5–10% of the population and resulting in mild elevations in bilirubin.²⁰ In these cases, it may be appropriate for more than one third (37.9%, 145/528) of these patients to not get follow up testing. Lastly, inertia may also play a critical role. Well documented in the management of hypertension and diabetes, clinical inertia can occur in all phases of care delivery, particularly with adherence to clinical guidelines.^21,22 The absence of timing recommendations in current guidelines and few clear links between liver test abnormalities and specific liver disease may leave clinicians with uncertainty or ambivalence when confronted with liver test abnormalities. Regardless of the underlying reasons, our data suggest that clinicians and patients don’t pursue rapid reassessment, resulting in large proportions of patients with abnormalities going more than 1 year without repeat.

How primary care clinicians view the utility of liver tests likely plays a significant role in follow-up. Given the frequency of abnormality, limited specificity, and limited evidence on the relationship between abnormality and outcomes, physicians may not see abnormal liver tests as clinically meaningful signals. The lower likelihood of repeat testing in patients with original abnormalities coming from a PCMH provider emphasizes this point. Certain factors appeared to intensify the signal, including the number of liver tests abnormal on a given panel (Figure 1), and the degree of abnormality (Figure 2). Clinicians may infer a higher likelihood of disease with a greater number and degree of liver test abnormality, leading to more follow-up testing. Another consideration is whether or not these repeat studies constitute follow-up at all. With the delays in testing and the association with multiple comorbidities, these “follow-up” tests may merely constitute comprehensive routine testing with patient visits, or continued monitoring for chronic conditions and medications. If, as some evidence suggests, abnormal liver tests are clinically meaningful, these follow-up practices may be inadequate. But, if abnormal liver tests are not diagnostically useful, then these labs are obtained too often. Though limited to ICD-9/10 codes using the Elixhauser comorbidity index in this study, few patients 355 (3.7%, 355/9545) patients in the sample received a liver-related diagnosis code. Given that the prevalence of NAFLD alone in the U.S. is anywhere from 20–30%, this small number of patients with liver-related diagnoses offers opportunity for improvement.²³ We speculate that clinician and patient perspectives on liver test abnormalities have broad implications on both over-testing and under-diagnosis.

We recognize limitations of this study. First, the study included data derived from a single center, perhaps limiting its generalizability. However, with the patient population belonging to a general internal medicine clinic, there is almost certainly broader generalizability, i.e., beyond that offered by a specialty clinic. Given the observational nature and reliance on administrative information, the data from this probably does not fully encapsulate the clinical circumstances under which the index testing was ordered. But with the varying reasons providers ascribe to ordering liver tests in the literature, this broad approach provides a more comprehensive picture of the information confronting clinicians in a PCMH. Also, this data doesn’t suggest the appropriate time to repeat testing. The only way to fully know the “right” time would require an understanding of the linkage of abnormal liver tests and diagnoses (liver-related and otherwise). Unfortunately, we only have access to ICD-9/10 codes for liver diagnosis at this time and ICD-9/10 coding has been shown to have limited agreement when compared to chart review for the diagnosis of liver disease (κ=0.50).²⁴ This study did not include other investigations (viral hepatitis serology, imaging, etc.) in follow-up to abnormal liver tests, which may underestimate the depth and frequency of investigation. Lastly, patients with follow-up liver tests done outside of the MUSC system, as well as abnormal tests prior to the study period, cannot be accounted for in the data set. Knowledge of such data may impact the patterns of follow-up observed.

Conclusions

Many patients seen in a primary care setting have abnormal liver tests and the presence and timing of repeat testing varies. The proportion of patients having follow-up tests increased over time, but many patients still lacked repeat testing at 1 year. The number of liver tests abnormal and degree of elevation were inversely associated with the time to repeat testing. We speculate that a better understanding of the clinical implications and the necessary follow-up for abnormal liver tests is needed to improve healthcare quality.