National Lung Screening Trial (NLST) investigators reported a 20% reduction in lung cancer mortality and a 6.7% reduction in all-cause mortality for participants screened with low-dose computed tomography (LDCT) compared with chest x-ray (CXR).1 Despite concerns voiced by a Medicare Evidence Development and Coverage Advisory Committee (MEDCAC) and the American College of Physicians (ACP),2,3 the United States Preventive Services Task Force assigned a grade B rating to LDCT lung screening, ensuring that lung screening is a covered service under the Affordable Care Act,4,5 and the Centers for Medicare and Medicaid Services (CMS) elected to cover LDCT lung screening.6 Costs, and in particular potential costs associated with the relatively high rate of significant incidental findings (SIFs), were cited as concerns by MEDCAC and the ACP.2,3
SIFs are findings that are incidental to the reason that the test is ordered, but are of potential medical significance, and have been noted in as many as 14%,7,8 of patients undergoing LDCT lung screening. SIF detection has been associated with costs for diagnostic testing and treatment.9,10 Although costs for LDCT versus CXR screening were expected to be higher in the NLST cost-effectiveness analysis (CEA)11 due to the high rate of false positive screens (∼25% per screen), at the time that the study was published and when the CEA was performed, the impact of SIFs on costs related to LDCT screening was not known. The NLST CEA, which estimated that LDCT lung screening cost $81,000 per quality-adjusted life year saved, assumed a cost of $500 to manage each participant with at least 1 SIF.11 SIF detection may lead to the early diagnosis of cancer and cardiovascular conditions,8,9,12–16 but whether that early detection translates to reductions in patient morbidity and mortality is unknown, and, in addition to potential benefits of SIF detection, may be associated with iatrogenic morbidity and mortality, if the discovery of these findings stimulates a highly invasive diagnostic work-up for what are ultimately determined to be benign conditions.17
In this paper, we use Medicare fee-for-service claims data to determine the comparative costs across the NLST LDCT and CXR arms, and the impact of SIFs on that comparison, comparing total and diagnostic (radiologic) costs between LDCT and CXR screening arms for NLST participants aged 65 years and older. We examine the per participant and per screen total and diagnostic costs by screen results (positive with abnormalities suspicious for lung cancer, SIF, negative).
The NLST, a collaboration between the American College of Radiology Imaging Network (ACRIN), now part of the ECOG-ACRIN Cancer Research Group, and the Lung Screening Study has been described elsewhere.18 In brief, the multi-institutional trial of 53,452 participants was designed to determine whether LDCT reduced lung cancer mortality relative to CXR. Participants were scheduled to receive 3 screening examinations: a baseline (first) and 2 incidence (second and third) screens at 1-year intervals.
In total, 18,840 NLST participants were enrolled at 23 ACRIN sites and were asked to provide their social security number (SSN) at recruitment. For the current paper, we limited our study population to those 65 and older at accrual and examined data obtained through linkage with Medicare administrative billing data based on SSN facilitated by the CMS Research Data Assistance Center.
Data on Abnormalities From the NLST
Screening results were obtained from the ACRIN database through the ECOG-ACRIN Biostatistics and Data Management Center at Brown University. Screens were classified hierarchically as: (1) positive with abnormalities suspicious for lung cancer; (2) negative for lung cancer, but with significant abnormalities not suspicious for lung cancer (negative with SIFs); or (3) negative without SIFs. This last category included screens with minor abnormalities not suspicious for lung cancer and screens with no abnormalities. SIFs were further classified by anatomic location.
CMS patient-level data included the standard analytical files, which provided details about individual inpatient and outpatient visits, including dates of service, ICD-9 diagnosis and procedure codes, diagnosis-related group (DRG) codes, and CMS common procedure (HCPCS) codes.
Retrieval of CMS Data
Patients were eligible to enter the cohort if they were enrolled into the ACRIN arm of the trial, provided a valid SSN, for which we located a match in the Medicare fee-for-service data files, and had Medicare parts A and B data available.
We evaluated both total costs and radiologic costs following LDCT and CXR screening. We reasoned that total costs would be most comprehensive, including the costs of managing complications resulting from work-up and treatment for any associated conditions, as well as some costs completely unrelated to screening, whereas radiologic costs would be related more directly to the screening result. In estimating radiologic costs, we included costs for current procedural terminology (CPT) codes 7000-8000 (excluding codes relating to radiotherapy: 77261-77799), as we considered those most likely related to radiologic evaluation of screen findings.
Annual costs were estimated using outpatient procedures and hospitalizations documented in the Medicare claims and the dollar value Medicare would reimburse for the associated DRG and CPT codes. For inpatient care, costs for the year in which the medical care occurred were obtained from the “DRGPRICE,” item 43 in the MedPAR RIF file (www.resdac.org/cms-data/files/medpar-rif/data-documentation). For outpatient care, costs were calculated using CPT costs based on the Physician Fee Schedule, including both professional and technical components. For part B, if a visit was reported with an associated CPT code, but there was no cost available for that CPT in the Physician Fee Schedule, then the cost for that visit was imputed from a regression model using other CPT costs for which payment was available, with adjustment for patient sex and age. Because reimbursements for positron emission tomography/computed tomography are set locally, we assigned a value of $1284 based on Medicare payments to Dartmouth-Hitchcock Medical Center in 2009. If a patient had an outpatient visit reported without an associated CPT code, then we used the actual payment amount for the cost of that visit. In such cases, these costs were not classified as radiologic.
The costs of screening LDCT and chest x-ray were borne by the trial. We did not include those costs in our analyses, as they were not billed to Medicare, and our goal was to examine the downstream costs associated with screening.
We did not adjust for inflation because we focused on comparative costs and the time period examined was short (limited to 3 years).
We compared participants for whom CMS data were and were not available with respect to baseline patient characteristics to determine whether patients included in our analyses differed in any way from those for whom we were unable to obtain cost data.
To assess the impact of SIFs on costs, we used 2 analytic approaches: a participant-level analysis and a screen-level analysis. For the participant-level analyses, we compared the average annual total costs and radiologic costs by randomized study arm (LDCT vs. CXR) and according to screening outcomes. The screening outcome analyses classified participants into the following mutually exclusive categories: ever positive, meaning the participant had at least 1 screening result positive with abnormalities suspicious for lung cancer; ever SIF, meaning the participant had no screening results positive for lung cancer, but at least 1 negative result with a SIF; and always negative, meaning that all screens were negative without SIFs. Costs were measured from the time of the first screen until 1 year after the third screen. In the participant-level analyses, if a participant died during the course of the trial or was not screened in any given year, the costs that participant incurred in the CMS data for that year were included. We present the data as annual averages, rather than the total 3-year cost for comparability with the screen-level analyses. We estimated the relative difference in cost for ever positive and ever SIF categories compared with always negative. For this approach, we used generalized linear models adjusted for sex, age, arm of the trial, result category, and the interaction of result category by arm, with a log link and a quasi-likelihood assumption, in which the variance is proportional to the mean.
In the screen-level analyses, we examined the costs for participants with a “positive” screen with those for participants with a “negative with SIFs” screen and participants with a “negative without SIFs” screen at each of the 3 screening timepoints. In the screen-level analyses, participants were included for any year that they received a screening test, even if they died in that 1-year period. We defined the year following the screen as the time until the next screen or 1 year, whichever came sooner. A generalized estimating equation approach adjusted for sex, age, arm of the trial, result category, and all possible interactions between result category, screening timepoint, and arm was used treating the screen-specific costs as repeated measures with a compound symmetry working covariance matrix and a log link, because participants could have been screened up to 3 times. Relative costs were compared across screen results and over time using Wald tests. All analyses were conducted using SAS PROC GENMOD.
To determine whether cost differences were attributable to the larger number of lung cancers detected in the LDCT screening arm, we repeated the participant-level and screen-level analyses excluding participants diagnosed with lung cancers during the course of the trial.
Of 5216 NLST participants who were 65 and older at the first screening examination, 2610 were in the LDCT arm and 2606 in the CXR arm. In total, 42% of the computed tomography (N=1100) and 50% of the CXR (N=1294) had data available for analysis after excluding those without a valid SSN or who were HMO participants during the study period (Fig. 1). The LDCT and CXR SSN cohorts with CMS data available were similar with respect to the characteristics described in Table 1. Comparisons of our study population and the remaining NLST population are included in the online Appendix (Supplemental Digital Content 1, http://links.lww.com/MLR/B547).
There were substantially more SIFs in the LDCT than in the CXR study arm (Table 2) with cardiovascular abnormalities being most common for both arms.
In the participant-level analysis (Table 3), total costs did not differ significantly between the LDCT and CXR arms ($11,029 vs. $10,905, respectively, P=0.85; 95% CI for percentage LDCT/CXR, 90%–113%). There were also no statistically significant differences in costs between the LDCT and CXR arms within screen result categories (ever positive, ever SIF, always negative). Similarly, radiologic costs did not differ significantly across arms (LDCT, $1085; CXR, $1062; P=0.68), nor did these costs differ within screen result categories. Although not significantly different, the costs were higher for ever SIFs in the LDCT arm than the CXR arm ($13,744 vs. $9,669; P=0.1090), whereas for always negatives, costs were lower in the LDCT versus the CXR arm ($9732 vs. $10,875; P=0.1479).
When we repeated analyses excluding lung cancers (because more lung cancers were diagnosed and treated in the LDCT arm), results were similar. Total costs were $11,132 for LDCT versus $10,913 for CXR screen participants (P=0.74). Radiology costs were $1082 for LDCT versus $1062 for CXR screen participants (P=0.73).
In the screen-level analyses (Table 4), LDCT and CXR arm screen costs for SIF or negative screens were not statistically significant different at any timepoint. Costs were statistically significant higher for LDCT versus CXR screen positive patients at the second screen, but not at the first or third screens. The same pattern was seen for radiology costs.
We found no significant difference in average annual costs (total and radiologic) for participants screened with LDCT compared with CXR in the ACRIN/NLST fee-for-service Medicare population, despite the higher number of SIFs in the LDCT arm of the trial. Our comparison across randomized arms of the study strengthens our ability to draw causal inference regarding the comparative costs following screening with LDCT and CXR and provides evidence that the higher level of SIF detection at LDCT screening may not increase costs of LDCT as opposed to CXR screening. The similarity in costs across arms is especially notable because these comparisons were made in the older Medicare-eligible NLST participants, who are generally assumed to have the highest rates of comorbidities.17
The lack of difference in total costs between the LDCT and CXR arms is potentially significant. The higher costs associated with the high rate of ever SIF findings in the LDCT arm of the study ($13,744, N=200) may be offset by reduced costs in LDCT patients who consistently screened negative ($9732, N=576). In contrast, costs for CXR participants who consistently screened negative were high and the number of participants in this category was also large ($10,875, N=1019). Whether the higher costs incurred by participants with ever SIF results in the LDCT arm were associated with earlier disease detection and a morbidity or mortality benefit is unknown. Further study is needed to investigate whether this was the case.
Our findings suggest that the differential cost of lung cancer screening with LDCT, versus CXR, may have been less than previously estimated in the NLST CEA. In the NLST CEA, the estimated difference in per person cost between the LDCT and CXR arms was about $444 (Appendix, Supplemental Digital Content 2, http://links.lww.com/MLR/B548). Spread over the 3-year screening period, this is about $148 per year for participants aged 55–74 at the time of screening initiation, after exclusion of screening examination costs,11 which were excluded from our current analyses. In comparison, we estimated the difference in per person cost to be $124 (95% CI, $1128–$1376), despite the older age of our Medicare population. The fact that costs did not significantly differ across the screening arms is reassuring, because information on actual overall medical costs was not available to NLST CEA investigators who relied on assumptions of SIF costs. Our analysis of Medicare claims data was based on actual costs and did not require such assumptions and suggests that lung cancer screening with LDCT, in comparison with no screening, may be slightly less costly than previously estimated. Although discounting was applied in the NLST CEA conducted using within-trial (6–8 y including the follow-up period) and lifetime horizons, we did not feel that it was necessary in our work, as discounting would have made little difference over the 3-year span under study.
In the screen-level analyses, at the second screen, the costs were significantly lower for participants with positive screens following LDCT versus CXR screening. This may have been due to costs associated with the detection of more advanced lung cancers in the CXR arm of the trial.
For this study we compared costs accrued in the year following a screening result. We have no evidence that these costs were directly related to the screening tests, but given the randomized design, we assume that contrasts between trial arms were unlikely to be affected by confounding with preexisting comorbidities or by other factors that might affect costs, which were equally distributed across trial arms by the randomization process. We are reassured that our results were not affected by patient comorbidities, as our results were unchanged when we adjusted for baseline patient-reported comorbidities, and the screening process for inclusion in the NLST precluded the enrollment of participants with serious comorbidities.
Although the sample size available for this study is relatively small, the confidence limits are fairly tight. Postscreening costs for those screened using LDCT seem to be between 10% less and 13% more than for those screened using CXR. Furthermore, our data provide the only randomized, trial-related, cost outcomes for LDCT screening relative to CXR. The randomized design of this study strengthens our ability to draw causal inferences with respect to cost comparisons.
CXR screening does not reflect standard of care, however, costs seen in the CXR arm of the study may be assumed to approximate those that would be seen in an unscreened comparison group as was done in the NLST CEA.11 Furthermore, Oken et al19 found no significant benefit of CXR as opposed to no screening. As well, though we report CXR SIFs, the nature of SIFs detected by LDCT and CXR are different.
Our estimates were based on medical services received by fee-for-service Medicare patients. Our goal in these analyses was to examine relative costs of screening with LDCT and CXR (as a proxy for no screening). We acknowledge that the actual dollar costs may vary for patients screened for lung cancer who were enrolled in Medicare managed care plans or in other insurance plans. We have no reason, however, to believe that the relative costs will differ based on insurer. We feel that providing information on the costs associated with LDCT screening is important for insurance company and Medicare planning, even if screening is a covered service.
Finally, in this paper we focused on costs associated with SIF management. We did not examine potential benefits of these findings. The NLST was the first cancer screening study to report a reduction in all-cause mortality, and nearly one third of the reduction was due to a reduction in deaths from other causes. Further work is needed to elucidate the impact of SIFs on patient outcomes.
We found little difference in total or diagnostic costs between the LDCT and CXR arms of the NLST in this analysis using Medicare data, despite the higher number of SIFs in the LDCT arm of the study.
The authors thank Loretta H. Pearson for her work coordinating this project and Pratikkumar Desai, MPH and Erin Greco, MS, formerly with the American College of Radiology Biostatistics Center, Brown University, Providence, RI for their assistance with the data analysis. The authors also thank the Screening Center investigators and staff of the National Lung Screening Trial (NLST). Most importantly, the authors acknowledge the study participants, whose contributions made the NLST possible.
1. National Lung Screening Trial Research Team, Aberle DR, Adams AM, Berg CD, et al. National Lung Screening Trial Research Team. Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med. 2011;365:395–409.
2. MEDCAC Meeting. Lung Cancer Screening with Low Dose Computed Tomography. 2014. Available at: https://www.cms.gov/medicare-coverage-database/details/medcac-meeting-details.aspx?MEDCACId=68
. Accessed March 15, 2018.
3. Owens DK, Qaseem A, Chou R, et al. Clinical Guidelines Committee of the American College of Physicians. High value, cost-conscious health care: concepts for clinicians to evaluate the benefits, harms, and costs of medical interventions. Ann Intern Med. 2011;154:174–180.
4. Health Policy Brief: Preventive Services Without Cost Sharing, Health Affairs
, December 28, 2010.
5. Patient Protection and Affordable Care Act, 42 U.S.C. § 18001 (2010). Available at: www.gpo.gov/fdsys/pkg/PLAW-111publ148/content-detail.html
. Accessed August 28, 2013.
6. Decision memo for screening for lung cancer with low dose computed tomography (LDCT) (CAG-00439N). Available at: https://www.cms.gov/medicare-coverage-database/details/nca-decision-memo.aspx?NCAId=274
. Accessed March 15, 2018.
7. Jacobs PC, Mali WP, Grobbee DE, et al. Prevalence of incidental findings in computed tomographic screening of the chest: a systematic review. J Comput Assist Tomogr. 2008;32:214–221.
8. Swensen SJ, Jett JR, Sloan JA, et al. Screening for lung cancer with low-dose spiral computed tomography. Am J Respir Crit Care Med. 2002;165:508–513.
9. Kucharczyk MJ, Menezes RJ, McGregor A, et al. Assessing the impact of incidental findings in a lung cancer screening study by using low-dose computed tomography. Can Assoc Radiol J. 2011;62:141–145.
10. Priola AM, Priola SM, Giaj-Levra M, et al. Clinical implications and added costs of incidental findings in an early detection study of lung cancer by using low-dose spiral computed tomography. Clin Lung Cancer. 2013;14:139–148.
11. Black WC, Gareen IF, Soneji SS, et al. National Lung Screening Trial Research Team. Cost-effectiveness of CT screening in the National Lung Screening Trial. N Engl J Med. 2014;371:1793–1802.
12. MacRedmond R, McVey G, Lee M, et al. Screening for lung cancer using low dose CT scanning: results of 2 year follow up. Thorax. 2006;61:54–56.
13. van de Wiel JC, Wang Y, Xu DM, et al. NELSON study group. Neglectable benefit of searching for incidental findings in the Dutch-Belgian lung cancer screening trial (NELSON) using low-dose multidetector CT. Eur Radiol. 2007;17:1474–1482.
14. Rampinelli C, Preda L, Maniglio M, et al. Extrapulmonary malignancies detected at lung cancer screening. Radiology. 2011;261:293–299.
15. Chiles C, Paul NS. Beyond lung cancer: a strategic approach to interpreting screening computed tomography scans on the basis of mortality data from the National Lung Screening Trial. J Thorac Imaging. 2013;28:347–354.
16. Chiles C, Duan F, Gladish GW, et al. NLST Study Team. Association of coronary artery calcification and mortality in the National Lung Screening Trial: a comparison of three scoring methods. Radiology. 2015;276:82–90.
17. Humphrey L, Deffebach M, Pappas M, et al. Screening for Lung Cancer: Systematic Review to Update the US Preventive Services Task Force Recommendation Evidence Synthesis No 105 AHRQ Publication No 13-05188-EF-1. Rockville, MD: Agency for Healthcare Research and Quality; 2013.
18. National Lung Screening Trial Research Team, Aberle DR, Berg CD, Black WC, et al. National Lung Screening Trial Research Team. The National Lung Screening Trial: overview and study design. Radiology. 2011;258:243–253.
19. Oken MM, Hocking WG, Kvale PA, et al. Prostate, Lung, Colorectal, and Ovarian (PLCO) Project Team. Screening by chest radiograph and lung cancer mortality: the Prostate, Lung, Colorectal, and Ovarian (PLCO) randomized trial. JAMA. 2011;306:1865–1873.