Gaps in Guideline-Concordant Use of Diagnostic Tests Among Lung Cancer Patients

Abstract

Background

Practice guidelines recommend routine use of pulmonary function tests (PFTs), computed tomography (CT), and positron emission tomography (PET) for the workup of resectable lung cancer patients. Little is known about the frequency of guideline concordance in routine practice.

Methods

A cohort study (2007 to 2013) of 15,951 lung cancer patients undergoing lobectomy or pneumonectomy was conducted with MarketScan, a claims database of individuals with employer-provided health insurance. Guideline concordance was defined by claims for PFT within 180 days of resection and for CT and PET within 90 days of resection. Generalized linear models were used to evaluate temporal trends, patient characteristics, and costs associated with guideline-concordant care.

Results

Overall, 61% of patients received guideline-concordant care, increasing from 57% in 2007 to 66% in 2013 (p < 0.001). Compared with patients who received guideline-discordant care, patients with guideline-concordant care more frequently underwent repeat testing (PFT: 21% versus 12%, p < 0.001; CT: 46% versus 22%, p < 0.001; PET: 2.3% versus 1.1%, p < 0.001). Health plan–adjusted mean total test-related costs were higher among guideline-concordant patients who underwent repeat testing than patients who did not ($4,304 versus $3,454, p < 0.001).

Conclusions

Forty percent of lung cancer patients treated with surgical procedures did not receive recommended noninvasive cancer staging and physiologic assessment before resection. Guideline concordance was associated with repeat testing, and repeat testing was associated with higher costs. These findings support the need for quality improvement interventions that can increase guideline concordance while curbing potential excess use of diagnostic tests.

Variability in outcomes has led to concerns over the delivery of lung cancer care in the United States [1]. Efforts to bridge this quality gap largely consist of performance feedback with risk-adjusted short-term outcomes after pulmonary resection [2]. Interest is growing in measuring process compliance as a means of facilitating high-quality cancer care [3], but little is known about process-of-care use among lung cancer patients in the community at large. It is difficult to develop quality improvement initiatives or to predict their impact without an understanding of the current status of process compliance in routine practice. Furthermore, identification of patient and health system factors associated with compliance would inform the development and evaluation of quality improvement interventions and health policy.

There are two fundamental aspects of evaluating a patient with potentially operable lung cancer: (1) physiologic evaluation with a pulmonary function test (PFT) to determine the patient's candidacy for extent of resection and (2) staging with computed tomography (CT) and positron emission tomography (PET). These noninvasive tests inform optimal patient and treatment selection. Practice guidelines from the American College of Chest Physicians (ACCP) and National Comprehensive Cancer Network (NCCN) recommend routine use of PFT, CT, and PET among patients with potentially curable non-small cell lung cancer [4–7]. The goal of these guidelines is to reduce nontherapeutic interventions and to improve the accuracy of outcome reporting with standard staging. Compliance with practice guidelines is expected to be high (approaching 100%) among patients selected to undergo resection, particularly among patients undergoing lobectomy and pneumonectomy. The goal of this investigation was to characterize the frequency of and factors and costs associated with guideline-concordant use of noninvasive diagnostic tests before pulmonary resection.

Material and Methods

We conducted a retrospective study of adult lung cancer patients who underwent lobectomy or pneumonectomy between July 1, 2007, and December 31, 2013, with the use of MarketScan (Truven Health Analytics, Ann Arbor, MI). This database contains inpatient and outpatient claims for more than 35 million insured employees and their dependents in the United States [8]. The University of Washington Institutional Review Board waived the need for consent and approved this investigation. The cohort was defined by enrollees with an International Classification of Diseases, Ninth Edition (ICD-9) diagnostic code for lung cancer and an ICD-9 procedure and/or common procedure terminology code for lobectomy or pneumonectomy (Appendix). To ensure complete ascertainment of diagnostic test use and therapy, continuous health plan enrollment was required from 180 days before and through the end of the index hospitalization. Patients were sequentially excluded for the following reasons: missing end enrollment date (n = 1), less than 180 days of continuous enrollment before resection (n = 3,149), no continuous enrollment through the end of their index hospitalization (n = 2,718), missing health plan information (n = 632), those who received neoadjuvant radiation therapy or chemotherapy (n = 2,285), and negative or zero values for provider and facility claims cost data (n = 863). The final study cohort consisted of 15,951 patients.

Guideline concordance was defined by at least one claim for PFT, CT, and PET before resection. NCCN and ACCP practice guidelines relevant to the study period recommended use of PFT, CT, and PET and were invariant in their recommendations between 2007 and 2013 for the use of these tests [4–7, 9]. For PFT, a 180-day period of test ascertainment was chosen on the basis of the clinical plausibility of obtaining a meaningful and informative result, a precedence for using this time period in the literature [10], and because it is used as a quality improvement metric in the Lung Cancer ProvenCare initiative [3]. For CT and PET, a 90-day period of ascertainment was chosen to avoid measuring noninvasive imaging performed for reasons other than staging, the most common being a short 3-month interval follow-up for the detection of a new pulmonary nodule [11]. Use of diagnostic tests was ascertained with all available information within MarketScan (Truven Health Analytics), including ICD-9 procedure, common procedural terminology, revenue center, service subcategory, and procedure group codes (Appendix). Patients were considered to have received a PET if they underwent a stand-alone PET, fusion PET/CT, or integrated PET/CT.

Other variables in this study included age, sex, modified Charlson comorbidity index [12], extent of resection, health plan type, inpatient death, repeat use, and test-related costs. Repeat use of PFT, CT, and PET were defined by a claim(s) on a date subsequent to the first claim within the prespecified test ascertainment interval. The cost of each initial diagnostic test was the sum of reimbursements for provider and facility claims on the first date in the ascertainment interval. Analyses of total costs included the sum of all reimbursements for initial and repeat testing within the ascertainment interval. Because MarketScan (Truven Health Analytics) contains information on copayments and deductibles, the sum of these payments were reported as out-of-pocket patient expenses. The analysis is from the patient and payer perspectives. Costs were adjusted for inflation with the use of the Medical Consumer Price Index [13].

STATA/SE version 12.1 (StataCorp LP, College Station, TX) was used for all analyses. Categorical variables were compared with the χ² test. For continuous variables, medians were compared with the Kruskal-Wallis test, and means were compared with a t test that allowed for unequal variance. Multivariable regression was used to evaluate temporal trends in and factors associated with guideline-concordant care. All models were adjusted for age, sex, comorbidity index, type of resection, and health plan. Because guideline concordance was not expected to be a rare event, we estimated relative risks with the use of generalized linear models with a log-link function, specifying a Poisson distribution and clustering on health plan type [14]. Cost models were adjusted for health plan type with the use of generalized linear models. A log-link function was selected for all models and was tested with the Pregibon link test (all models p > 0.05) and a modified Hosmer-Lemeshow goodness of fit test (all models p > 0.05). Modified Park tests were used to select the family function that best modeled the mean-variance data structures. For initial and total cost models the Poisson family was used; for out-of-pocket cost models the gamma family was selected [15]. All cost models were adjusted for health plan. Two-sided p values of less than 0.05 were considered statistically significant.

Results

Among 15,951 lung cancer patients undergoing operations (Table 1), guideline-concordant use of diagnostic tests (at least one PFT, CT, and PET within the pre-specified time frame) occurred in 61% (95% confidence interval [CI]: 60% to 62%). This proportion increased from 57% in 2007 to 66% in 2013 (p-trend < 0.001) (Fig 1). Over the same time period the proportion of women increased, fewer patients underwent pneumonectomy, and more patients had comorbid conditions (all p-trends < 0.001). The increasing trend in guideline concordance remained significant after adjusting for changing cohort characteristics over time (adjusted p-trend < 0.001). The most common reason for guideline discordance was the absence of a claim for PET within the prespecified time frame, followed by the absence of a claim for PFT and then CT.

Table 1. Cohort Characteristics According to Guideline Concordance.

	All Patients (n = 15,951)	Nonconcordant (n = 6,156)	Concordant (n = 9,795)	p Value
Mean age (SD), years	64 (11)	64 (11)	64 (11)	<0.001
Male, %	48	49	47	0.04
Charlson comorbidity index, %				0.003
0	49	51	48
1	39	37	40
2	9.1	9.0	9.1
≥3	2.9	3.0	2.8
Pneumonectomy, %	7.4	7.1	7.7	0.19
Health plan, %				<0.001
PPO	54	52	55
Comprehensive	23	22	23
HMO	12	14	11
POS	6.9	7.1	6.7
HDHP/CDHP	3.1	3.3	3.1
EPO	1.2	1.1	1.3
Inpatient death	1.3	1.5	1.1	0.03

CDHP = consumer-directed health plan; EPO = exclusive provider organization; HDHP = high-deductible health plan; HMO = health maintenance organization; POS = point of service; PPO = preferred provider organization; SD = standard deviation.

Fig 1.

Temporal trends in guideline-concordant use of noninvasive diagnostic tests and individual diagnostic modalities. (CT = computer tomography; PET = positron emission tomography; PFT = pulmonary function test.)

Patients who received guideline-concordant care were more frequently women, older than age 50 years, and had a comorbidity index of 1 than patients who received nonconcordant care (Table 1). No difference was found according to procedure. These associations persisted in multivariable analysis (Table 2). Adjusted rates of guideline adherence varied across health plans. The greatest disparity in rates of guideline concordance according to health plan type was between preferred provider organizations in which patients were most likely to receive guideline-concordant care, and health maintenance organizations (HMOs) in which patients were least likely to receive guideline-concordant care. Guideline concordance among patients enrolled in high-deductible/consumer-directed health plans (HDHPs/CDHPs) was not significantly different than any other health plan with the exception of HMO, whereby patients were 10% less likely to receive guideline-concordant care (relative risk 0.90, 95% CI: 0.83 to 0.97, p = 0.01).

Table 2. Factors Associated With Receipt of Guideline-Concordant Care.

	RR (95% CI)	p Value
Age, years
<30	0.37 (0.30–0.46)	<0.001
30–39	0.59 (0.57–0.61)	<0.001
40–49	0.91 (0.88–0.93)	<0.001
50–59	Reference
60–69	1.0 (0.97–1.03)	0.839
≥70	1.0 (0.94–1.06)	0.962
Male sex	0.97 (0.94–0.99)	0.008
Comorbidity index
0	Reference
1	1.04 (1.02–1.06)	<0.001
2	1.01 (0.99–1.03)	0.48
≥3	0.98 (0.92–1.05)	0.52
Pneumonectomya	1.06 (1.00–1.13)	0.06
Health plan
PPO	Reference
Comprehensive	0.98 (0.96–1.01)	0.26
HMO	0.86 (0.86–0.87)	<0.001
POS	0.95 (0.95–0.96)	<0.001
HDHP/CDHP	0.95 (0.88–1.03)	0.20
EPO	1.03 (1.03–1.04)	<0.001
Year of diagnosis
2007	Reference
2008	1.02 (0.96–1.08)	0.62
2009	1.05 (1.00–1.11)	0.04
2010	1.07 (1.01–1.13)	0.03
2011	1.08 (1.04–1.14)	<0.001
2012	1.10 (1.07–1.14)	<0.001
2013	1.15 (1.13–1.17)	<0.001

^aCompared with lobectomy.

CDHP = consumer-directed health plan; CI = confidence interval; EPO = exclusive provider organization; HDHP = high-deductible health plan; HMO = health maintenance organization; POS = point of service; PPO = preferred provider organization; RR = relative risk.

The median time interval between initial test and resection varied by concordance (Table 3). The largest difference was 14 days for PFT. Patients who received guideline-concordant care more often underwent repeat PFT, CT, and PET before resection (all p < 0.001). The median total number of noninvasive diagnostic tests was significantly higher in the guideline-concordant group than in the nonconcordant group (four versus two tests, p < 0.001).

Table 3. Use of Noninvasive Diagnostic Tests According to Guideline Concordance.

	Nonconcordant (n = 6,156)	Concordant (n = 9,795)	p Value
Median days between initial test and resection
PFT	52	38	<0.001
CT	44	49	<0.001
PET	37	32	<0.001
Repeat use, %
PFT	12	21	<0.001
CT	22	46	<0.001
PET	1.1	2.3	<0.001

CT = computed tomography; PET = positron emission tomography; PFT = pulmonary function test.

The median costs of initial PFT and PET were slightly higher in the guideline-concordant group, but the costs of initial CT were not different (Table 4). The median out-of-pocket cost for all initial tests was $0, with mean unadjusted and adjusted out-of-pocket costs ranging from $8 to $58. Mean out-of-pocket costs accounted for a small fraction (1% to 7%) of total costs. After adjusting for health plan type, the mean total costs for tests performed varied as expected: guideline discordant care without repeat testing ($1,617), guideline discordant care with repeat testing ($2,601), guideline-concordant care without repeat testing ($3,454), and guideline-concordant care with repeat testing ($4,304). Among patients who received guideline-concordant care, patients who underwent repeat testing have significantly higher costs than patients who did not (p < 0.001).

Table 4. Costs of Initial Noninvasive Diagnostic Testing According to Guideline Concordance.

Cost of Initial Testa	Variable	Nonconcordant	Concordant	p Value
PFT (n = 13,417)	No. of patients	3,622	9,795
Totalb, $	Median [IQR]	163 [52–294]	168 [54–292]	0.07
	Mean ± SD	243 ± 297	259 ± 332	0.006
	Adjustedc mean (95% CI)	242 (236–249)	258 (255–263)	<0.001
Out-of-pocketd, $	Mean ± SD	8 ± 38	9 ± 45	0.36
	Adjustedc mean (95% CI)	8 (7–10)	9 (8–9)	0.62
	Fraction of mean total cost, %	3	3
CT (n = 13,830)	No. of patients	4,095	9,795
Totalb, $	Median [IQR]	523 [299–1,057]	517 [314–1,070]	0.36
	Mean ± SD	834 ± 972	832 ± 916	0.97
	Adjustedc mean (95% CI)	830 (779–881)	833 (812–855)	0.93
Out-of-pocketd, $	Mean ± SD	51 ± 171	58 ± 187	0.04
	Adjustedc mean (95% CI)	49 (47–52)	58 (55–62)	<0.001
	Fraction of mean total cost, %	6	7
PET (n = 12,441)	No. of patients	2,646	9,795
Totalb, $	Median [IQR]	1,820 [1,144–3,252]	1,935 [1,160–3,366]	0.04
	Mean ± SD	2,313 ± 1,795	2,377 ± 1,784	0.11
	Adjustedc mean (95% CI)	2,323 (2,270–2,376)	2,374 (2,358–2,389)	0.14
Out-of-pocketd, $	Mean ± SD	31 ± 170	28 ± 160	0.34
	Adjustedc mean (95% CI)	31 (24–38)	28 (25–31)	0.41
	Fraction of mean total cost, %	1.3	1.2

^aInitial test refers to the first recorded test within the specified time frame for each test (PFT, 180 days; CT and PET, 120 days).

^bIncludes gross payments, copayments, and deductibles (professional and facility fees combined).

^cMultivariable adjustment for health plan and year.

^dIncludes copayments and deductibles (professional and facility fees combined).

CI = confidence interval; CT = computed tomography; IQR = interquartile range; PET = positron emission tomography; PFT = pulmonary function test.

Comment

ACCP and NCCN practice guidelines recommend routine use of PFT, CT, and PET in the preoperative evaluation of patients undergoing pulmonary resection for lung cancer. There is rarely a reason to omit these tests in patients undergoing lobectomy or pneumonectomy. In a nationally representative cohort of working Americans and their dependents, we report that only 61% of non-small cell lung cancer patients treated with lobectomy or pneumonectomy underwent PFT, CT, and PET within specified time frames. Equally surprising was the finding that guideline-concordant care was associated with a greater likelihood of repeat testing with subsequent higher costs.

It is unclear why 39% of patients did not receive all three tests before resection in this study, although lack of evidence to link outcomes with timing of preoperative testing is one possibility. In this context, it is reasonable to question the importance of performing these tests within guideline-specified time frames [16]. However, in the absence of evidence it is also reasonable to follow expert consensus derived from specialty providers to define timeliness, one of the seven pillars of quality advanced by the Institute of Medicine [17]. Access to care may be another explanation for guideline discordance. We did not find any meaningful differences in the interval between initial use of a diagnostic test and resection across guideline-concordant and -discordant groups, suggesting that systematic delays between preoperative evaluation and treatment do not explain guideline-discordant care in this study. Another component of access to care involves the ability to see specialist physicians. Even in insured populations, there is variability in obtaining this level of care [18], and thoracic surgical oncologists are more likely to order comprehensive preoperative tests than general surgeons [1]. One limitation of this study is that MarketScan (Truven Health Analytics) does not have information on health systems, hospitals, surgeons, and other physicians and is limited to employer-based sources of medical insurance. Therefore, it was not possible to explore the potential contributions of these factors. One unique aspect of this study was our ability to evaluate out-of-pocket expenses, which has the potential to influence testing. We found that out-of-pocket expenses accounted for only a small fraction (≤7%) of the total costs of a given test, and the greatest absolute dollar amount for a given test was less than $58 on average. Although it is possible that these costs are important barriers to accessing care for some lower-income Americans, it seems unlikely that out-of-pocket costs constitute a major determinant of guideline adherence.

Another strength of our data source is the ability to evaluate the potential impact of health plans on guideline-concordant use of diagnostic tests. Our multi-variable analysis showed that the greatest variation in guideline concordance was by health plan. Health care reform has led to concerns that HDHPs/CDHPs may undermine the receipt of optimal care, particularly for high-expenditure conditions such as cancer [19]. We found no evidence of a relation between guideline adherence and HDHPs, which is consistent with other studies of colorectal, breast, and cervical cancers that also reported no relation between health care use and HDHPs [19]. Interestingly, HMO was the health plan most likely to provide guideline-discordant care. This finding is noteworthy because HMOs are both the payer and provider of health services and may provide insights into the anticipated effects of accountable care organizations on cancer care delivery. This duality of HMOs requires optimization of both quality and costs (ie, value), but controversies can arise because quality is not easily defined. For example, although the routine use of PET is supported by five randomized trials reporting higher accuracy of staging, none of the studies reported a survival benefit attributable to PET [20–24]. If one were to only consider survival as a marker of quality, then the only rational conclusion is that PET has no value. Yet several trials found that better staging accuracy led to avoidance of surgical procedures, and its inherent risks and costs, in occult stage IIIB/IV disease patients who would not benefit from operative management. Viewed from this perspective, many would consider PET to be a high-value intervention. More detailed investigations of variation in care across health plans may reveal opportunities to improve care delivery and patient outcomes.

Another key finding from this study is the association of guideline concordance with greater health care use and higher costs. The fact that 39% of patients did not receive guideline-concordant care underlies this association. It is equally evident that the tendency to repeat tests also explains, in part, this association. What is unmeasured in this study is whether repeat testing represents appropriate care. For example, the quality of PFT results depends on patient cooperation. When PFT and functional status are discordant, it is reasonable to repeat PFT. The granularity of this data set precludes an understanding of the appropriateness of re-ordering PFT. Nevertheless, from a clinical perspective, repeating a CT in nearly one-half of all patients does not seem reasonable. It is more likely that providers and patients are driving this because of unfounded concerns about the quality of images or progression of disease or perceived challenges in acquiring and viewing outside films. To the extent that repeat testing is unnecessary, our findings suggest that guideline concordance alone is an imperfect measure of quality and value.

Our study has several limitations not previously mentioned. Our cohort is younger and healthier than other large populations of lung cancer patients, which raises concerns about generalizability. We view our study to be complementary to other sources of information (eg, Medicare,) that tend to underrepresent working Americans and their dependents. In addition, findings may not be generalizable to contemporary guidelines published by the ACCP; as of May 2013, the ACCP states that PET is not necessary for patients with peripheral stage IA tumors or ground glass opacities [6]. However, our findings remain generalizable to the most recent NCCN guidelines. Another limitation is that the use of claims to measure health care delivery may have resulted in an underestimation of guideline-concordant care. For example, it is possible that patients paid completely out of pocket for their diagnostic tests without reporting it to the insurer and/or providers did not submit claims. Some patients may have had supplemental insurance, resulting in the absence of observable claims. Another limitation is the lack of clinical detail in claims data, such as test results or other factors that can influence decision making. A recent qualitative study attempted to understand factors underlying guideline-discordant lung cancer care and identified a dynamic interplay between the patient, physician, family, clinical team, and level of evidence [25]. Without this information the reasons underlying guideline-concordant care cannot be determined conclusively. Unfortunately, no database contains this level of detail, hence the value of complementary quantitative and qualitative studies.

Our findings are relevant to ongoing quality improvement efforts in thoracic operations. For example, our study results motivate the adoption of process-of-care oriented interventions, such as ProvenCare Lung Cancer, a multifaceted intervention that involves clinical leadership, continuous performance feedback, a checklist, and stakeholder engagement [3]. The findings may also motivate The Society of Thoracic Surgeons General Thoracic Surgery Database to provide feedback on process compliance, particularly because The Society of Thoracic Surgeons General Thoracic Surgery Database is already measuring the use of PFT, CT, and PET. Our study also raises a cautionary note about simply measuring compliance without assessing overall use and potential excess use. In this regard, our study findings provide empirical evidence of the need for initiatives that curb the use of unnecessary tests and interventions such as the Choosing Wisely campaign (ABIM Foundation, Philadelphia, PA).