Incidental radiologic evidence of interstitial lung disease and time to pulmonary evaluation

Abstract

Background

Interstitial lung diseases (ILD) are relatively rare fibrotic lung diseases that can cause significant morbidity and mortality. Early diagnosis of ILDs is crucial to improve patient survival. Incidental radiologic findings of interstitial lung disease may provide an opportunity for early diagnosis. We sought to estimate the incidence of incidental interstitial lung disease findings on CT scans ordered for other reasons and quantify the time to pulmonary function test or pulmonary referral.

Methods

In this retrospective cohort study performed at a large metropolitan health system, radiology reports from CT scans of the neck, chest, abdomen/pelvis, and coronary arteries of adult patients were reviewed for documentation of ILD findings based on the pre-defined search terms. We also descriptively examined time from CT scan to pulmonary function testing or pulmonology referral. The main outcomes were incidence of ILD findings on CT scans and frequency of follow up testing.

Results

Of the 239,535 scans that were included, there were 25,219 initial scans with ILD findings (10.8%). Of those patients, only 16.1% had pulmonary function testing or pulmonary referral at one year, and 23.9% at 5 years. Follow up testing was more common if interstitial lung disease findings were present on CT chest compared to other scan types. The median time to PFT for all scan types was 59 days and the median time to pulmonary referral was 112 days.

Conclusion

In conclusion, incidental findings of interstitial lung disease are common on many different types of CT scans but infrequently lead to follow up pulmonary function testing or referral to a pulmonary specialist. Significant delays in time were noted for pulmonary referrals. This study raises awareness of delay in diagnosing ILD.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12931-025-03390-8.

Introduction

Interstitial lung disease (ILD) is a heterogenous group of disorders affecting lung parenchyma which are classified together due to similar clinical, radiologic and physiologic manifestations [1–3]. The diagnosis of ILD can be challenging due to such a diverse group of diseases each with a relatively low incidence; few physicians can gain enough expertise to familiarize themselves with ILD. Furthermore, many ILDs often have an insidious onset and nonspecific initial symptoms which may be attributed to deconditioning, old age or to an alternate diagnosis [4]. It is imperative to establish an ILD diagnosis in the early stages to impact the course of the disease with treatments. Actual diagnosis requires integration of patient characteristics, symptoms, characteristic radiologic and pathologic findings, and careful consideration of alternative etiologies. A multidisciplinary conference involving pulmonologists, pathologists and radiologists is the current gold standard for diagnosis of ILD [5–7].

A delay between initial clinical signs and symptoms and diagnosis of ILD is common [8]. In a survey of 600 registered members of the Pulmonary Fibrosis Foundation patient registry, 43% of respondents experienced a delay of one year or more between symptom onset and diagnosis of ILD, with nearly half of those experiencing a diagnostic delay of greater than 3 years [9]. While it is not known if respondents were initially seen at an academic vs. community center, nearly 70% of people reported that referral to a subspecialist with expertise in ILD was the most Important contributing factor to obtaining a clear diagnosis [9]. A delay in diagnosis of ILD also is associated with increased morbidity and mortality [9, 10]. Though initial symptoms can be vague and difficult to identify early on, radiologic evidence of ILD may be present prior to clinical symptoms [11] and may be a useful tool in the early diagnosis of ILD. A recent editorial has emphasized that lung cancer screening programs using low-dose CT provide an additional opportunity to identify incidental ILA, which may represent early ILD and carry prognostic implications [12].

The aims of this study were to determine the prevalence of incidental radiologic ILD findings on computed tomography scans ordered for other reasons at a tertiary academic center and determine the time from the index scan to additional pulmonary work up such as pulmonary function testing or pulmonary referral.

Methods

Study design

We conducted a retrospective cohort study at a large metropolitan health system associated with a tertiary academic center. Radiologist reports from CT scans of the neck, chest, abdomen and pelvis, and coronary arteries of adult patients between January 2014 and December 2019 were reviewed for documentation of ILD findings either in the narrative and/or impression of the report. Exclusion criteria were patients less than 18 years old, those with previously diagnosed ILD or cancer (as cancer diagnosis tends to trigger serial CT imaging, leading to favoring bias to detect interstitial findings), and those whose date of death was prior to the scan date (either because the imaging was completed prior to patient death but documentation was not finalized prior to date of death, or the patient was brain-dead at the time of the CT scan). Patient records were provided by the Information Exchange Services through the Clinical and Translational Sciences Institute Best Practices Informatics Consulting. Records were accessed through Epic Electronic Medical Record (EMR) System. The EMR database was queried for procedure codes associated with the relevant imaging studies. Radiology reports were reviewed for pre-defined search terms using Structured Query Language (SQL) wildcard characters to flag reports that returned with any of the pre-specified search terms. Pre-specified search terms were interstitial, interstitial lung disease, fibrosis/fibrotic, reticulation, traction bronchiectasis, honeycomb/honeycombing, inter-/intra-lobular septal thickening, ground glass, and mosaic attenuation.

Statistical analysis

Demographics were summarized for all first abnormal CT scans by scan type using descriptive statistics. To investigate the association between demographics and scan type, Analysis of Variance (ANOVA) was used for continuous variables, and Chi-square test was used for categorical variables. Outcomes were (1) pulmonary function test, (2) pulmonary referral, and (3) pulmonary function test or pulmonary referral. ILD findings and occurrence of outcomes were summarized for all first abnormal CT scans by scan type. The first abnormal CT scan per patient was used to determine time to pulmonary function test and pulmonary referral, if any. Diagnostic delay was calculated based on time in days between the index CT scan and pulmonary function test and time in days between the index scan and referral to a pulmonologist. If a patient did not have the event, they were censored at the last date of contact. For those who had the outcome, time to the outcome was descriptively summarized using median and range. Outcomes were summarized using cumulative incidence estimates and 95% confidence intervals (CI) using the cumulative incidence function. Gray’s test was used to compare cumulative incidence across scan types. Fine-Gray regression models for competing risks were used, and sub-distribution hazard ratios (HR) and 95% CIs were obtained for both unadjusted models and models adjusted for smoking status and age. All p-values reported are two-sided, with a significant threshold set at less than 0.05. Statistical analyses were performed using R (version 4.0.2, R Core Team) and SAS (version 9.4, SAS Institute Inc., Cary, North Carolina).

Results

Identification of cohort and patient demographics

A total of 916,054 CT scans were identified for review. After clearing duplicate orders associated with the same scan, there were 656,305 individual scans remaining. 409,557 CT scans were excluded for pre-existing diagnoses of ILD or cancer (any type). An additional 7213 scans were excluded: 7208 were in patients less than 18 years of age and 5 scans were identified with the date of death prior to the scan date (Fig. 1). Of the 239,535 scans that were included, there were 41,583 scans (17.4%) with ILD findings identified based on the pre-defined search terms [Table 1]. The most common type of CT scan obtained in the cohort was abdomen/pelvis CT (57.9%) while chest CT was the least common (1.3%). Chest CT scans were the most likely scan type to have ILD findings reported (36.9%), followed closely by neck CT scans (34.0%), then coronary CT scans (11.0%) and finally CT scans of the abdomen/pelvis (7.0%). Of the total scans with ILD findings, 48.2% of them commented on the ILD findings in both the narrative and the impression [Table 1]. The most common radiologic descriptions of ILD findings in this cohort were ground glass (59.1%), fibrosis (29.8%), and interstitial (22.7%) [Supplemental Table 1].

Fig. 1.

Flow diagram depicting the total number of CT scans identified for review, reasons for exclusion, and final number of included CT scans in the study

Table 1.

Descriptive summaries of incidental ILD findings by scan type for all scans

	All scans (N = 239,535)	Abdomen/Pelvis CT scans (N = 138,662)	Chest CT scans (N = 3,116)	Coronaries CT scans (N = 10,999)	Neck CT scans (N = 86,758)
ILD findings, n (%)
Yes	41,583 (17.4)	9695 (7.0)	1151 (36.9)	1207 (11.0)	29,530 (34.0)
No	197,952 (82.6)	128,967 (93.0)	1965 (63.1)	9792 (89.0)	57,228 (66.0)
ILD finding radiologic report, n (%)
No	197,952 (82.6)	128,967 (93.0)	1965 (63.1)	9792 (89.0)	57,228 (66.0)
Narrative only	5383 (2.2)	1451 (1.0)	128 (4.1)	123 (1.1)	3681 (4.2)
Impression only	16,138 (6.7)	5010 (3.6)	399 (12.8)	571 (5.2)	10,158 (11.7)
Both narrative and impression	20,062 (8.4)	3234 (2.3)	624 (20.0)	513 (4.7)	15,691 (18.1)

Of the 41,583 scans with incidental radiologic ILD findings, 25,219 were the first CT scans to report ILD for those patients. The remaining 16,364 scans were performed on patients who had ILD already identified in earlier CT scans. The demographics of patients with their first abnormal CT scan identifying ILD findings are summarized in Table 2. This was examined to see if there was a higher incidence of ILD-related findings among advanced age and smoking status, which are known risk factors for IPF, the most common ILD, and some other ILDs. The mean age was 61.5 years, 52.2% were female, and 98.5% identified as non-Hispanic/Latino. Of those with documentation on smoking status (21,140 out of 25,219 patients), 68.4% were current or former smokers. Approximately 23% of patients were deceased at the end of the study period. There was a significant association between type of scan and smoking status, with those patients who had an abdomen/pelvis CT being more likely to have never smoked compared to other scan types (p < 0.001). Finally, there was a significant association between scan type and age, in that patients with ILD findings on CT abdomen/pelvis were younger as compared to other scan types (p = 0.004). There was no significant association between scan type and sex or ethnicity (Hispanic/Latino vs. non-Hispanic/Latino).

Table 2.

Comparison of demographics by scan type for first abnormal scans

Variable	First abnormal scans (N = 25,219)	Abdomen/Pelvis CT scans (N = 6,223)	Chest CT scans (N = 568)	Coronaries CT scans (N = 707)	Neck CT scans (N = 17,721)	P-value1
Sex, n (%)						0.758
Female	13,156 (52.2)	3254 (52.3)	294 (51.8)	355 (50.2)	9253 (52.2)
Male	12,062 (47.8)	2968 (47.7)	274 (48.2)	352 (49.8)	8468 (47.8)
Ethnicity, n (%)						0.656
Hispanic/Latino	367 (1.5)	99 (1.6)	6 (1.1)	9 (1.3)	253 (1.5)
Non Hispanic/  Latino	23,841 (98.5)	5916 (98.4)	541 (98.9)	686 (98.7)	16,698 (98.5)
Smoking status, n (%)						<0.001
Never smoked	6499 (30.7)	1803 (35.3)	145 (29.5)	161 (27.4)	4390 (29.4)
Secondhand  exposure	179 (0.8)	44 (0.9)	3 (0.6)	4 (0.7)	128 (0.9)
Smoker	14,462 (68.4)	3265 (63.9)	344 (69.9)	423 (71.9)	10,430 (69.8)
Age, mean (SD)	61.5 (17.2)	60.9 (19.0)	61.5 (15.2)	62.3 (17.6)	61.7 (16.6)	0.004

¹P-value is for Chi-square test for categorical variables and ANOVA for continuous variables

²Sex was missing for 1, ethnicity was missing for 1011, and smoking status was missing for 4079

Follow up with pulmonary function test or referral to a pulmonologist

Of all the patients with a first abnormal CT scan identifying ILD findings, 19.0% of them were followed up by either pulmonary function test or pulmonary referral during the study period; 14.2% went on to have a pulmonary function test after their CT scan and 9.0% were referred to a pulmonologist [Supplemental Table 2]. A total of 848 (3.4%) were lost to follow up. Median time to follow up was 1.3 (range 0–6.9.9) years.

Using descriptive statistics found in Supplemental Table 3 the median time to PFT for all scan types was 59 (range: 0–2310) days while the median time to pulmonary referral was 112 (range: 0–2363) days. The median time to either PFT or pulmonary referral was 42 (range: 0–2329) days. The delay was shortest if incidental ILD findings occurred on a CT chest: 46 (range: 0–1993) days to PFT and 31 (range: 0–1945) days to pulmonary referral. The delays to follow up testing were longest if incidental ILD findings occurred on a CT abdomen pelvis: 206 (range: 0–2202) days and 350 (range: 0–2165) days, respectively.

Table 3.

Hazard ratio estimates for pulmonary function test or pulmonary referral by scan type

	Unadjusted1						Adjusted2
	Pulmonary function test		Pulmonary referral		Pulmonary function test or pulmonary referral		Pulmonary function test		Pulmonary referral		Pulmonary function test or pulmonary referral
	HR (95% CI)	P-value1	HR (95% CI)	P-value1	HR (95% CI)	P-value1	HR (95% CI)	P-value1	HR (95% CI)	P-value1	HR (95% CI)	P-value1
Scan type		< 0.001		< 0.001		< 0.001		< 0.001		< 0.001		< 0.001
Abdomen/Pelvis CT scan	Ref		Ref		Ref		Ref		Ref		Ref
Chest CT scan	5.31 (4.29, 6.56)		2.93 (2.22, 3.87)		4.36 (3.62, 5.25)		5.69 (4.55, 7.11)		2.92 (2.18, 3.90)		4.55 (3.75, 5.53)
Coronaries CT scan	1.50 (1.11, 2.02)		2.16 (1.64, 2.83)		1.90 (1.56, 2.37)		1.54 (1.11, 2.13)		2.27 (1.71, 3.01)		2.01 (1.59, 2.53)
Neck CT scan	4.14 (3.67, 4.67)		2.60 (2.28, 2.95)		3.62 (3.28, 3.99)		4.43 (3.88, 5.05)		2.61 (2.28, 3.00)		3.77 (3.40, 4.19)

¹Fine- Gray regression model for competing risks was used to investigate the effect of scan type on Pulmonary function test, pulmonary referral or pulmonary function test/pulmonary referral. Sub-distribution hazard ratio and 95% CI are presented for each outcome for each scan type in comparison to abdomen/pelvis CT scan

²Adjusted for age and smoking status

Cumulative incidence estimates and HRs for the three outcomes (pulmonary function test, pulmonary referral and pulmonary function test or pulmonary referral) by scan type were calculated [Table 3; Fig. 2, Supplemental Table 4]. There was a significant effect of scan type on all outcomes with and without adjustment for smoking status and age (Table 3). Specifically, chest CT, neck CT and coronary artery CT were all significantly more likely to have pulmonary function testing or pulmonary referral compared to abdomen/pelvis CT scans. The unadjusted HR of having either pulmonary function test or pulmonary referral following incidental ILD findings on CT chest was 4.36 (95% CI: 3.63–5.25), whereas CT neck was 3.62 (95% CI: 3.28–3.99) and 1.90 (95% CI: 1.53–2.37) for CT coronary scan. Cumulative incidence of pulmonary function test or pulmonary referral varied by scan type (p < 0.001), with chest CT having the highest cumulative incidence (1 year: 23.1%, 95% CI: 19.6–26.9%), and abdomen/pelvis CT having the lowest incidence (1 year: 5.0%, 95% CI: 4.4–5.6%). These trends remained the same at the 2-year and 5-year estimates, with approximately 35% of those patients with incidental ILD findings on CT chest having either a pulmonary function test or a pulmonary referral at 5 years from index scan, whereas only 10% of those with CT abdomen/pelvis findings had either follow up ordered at 5 years.

Fig. 2.

a Cumulative incidence estimates for pulmonary function test using the cumulative incidence function. Gray’s test was used to investigate the effect of scan type on Pulmonary function test b Cumulative incidence estimates for pulmonary referral using the cumulative incidence function. Gray’s test was used to investigate the effect of scan type on pulmonary referral c Cumulative incidence estimates for pulmonary function test or pulmonary referral using the cumulative incidence function. Gray’s test was used to investigate the effect of scan type on pulmonary referral or Pulmonary function test.

Discussion

At a large metropolitan health system, the incidence of ILD findings on CT scans ordered for other reasons was high at 17.4%. However, follow-up investigation including either pulmonary function testing or pulmonary referral was uncommon: the cumulative incidence of either pulmonary function test or pulmonary referral at 1 year and 5 years was 16.1% and 23.9% respectively. The frequency of follow up testing or referral was higher if the abnormal imaging findings were noted on either a CT scan of the chest or neck compared to a CT scan of the coronary arteries or abdomen/pelvis. It is possible that the reason for ordering a CT scan of the chest or neck was more intimately related to respiratory symptoms and providers paid closer attention to the findings compared to CT coronary arteries or CT abdomen/pelvis where these incidental findings were not the focus.

Similarly, there was a significant association between type of scan and smoking status, with those patients who received a CT of the abdomen/pelvis being less likely to have ever smoked. We speculate this is because the other scan types investigate lungs, coronary arteries, and carotid arteries, which are directly affected by smoking.

Fine-Gray regression models show a significantly increased likeliness of appropriate follow up care after incidental findings were discovered on chest CT. While scans ordered on patients with known ILD were excluded, the reason for ordering a CT scan of the chest may be more related to respiratory concerns than other scan types, and ordering providers might be more likely to recognize and follow up on the incidental findings of ILD. Patients with incidental findings on abdomen/pelvis CT scans were least likely to have follow up testing. Also, abdomen/pelvis CT scans were less likely to have the incidental findings reported in both the narrative and impression of the CT report as compared to other scan types, making it easier for an ordering provider to miss.

The descriptive time to follow up testing for the population at risk also varied by scan type. The median time to PFT following incidental ILD findings on any scan type was 59 days; median time to pulmonary referral was 112 days; median time to either PFT or pulmonary referral was 42 days. Delay times were shortest when the scan type was CT chest and longest when the scan type was CT abdomen/pelvis, which correlates with our incidence data.

The window in which ILD findings appear on a CT scan but before a patient has symptoms may represent an important opportunity to identify the disease process early and connect the patient with the appropriate specialty care. On the other hand, access to specialty testing and care can be limited, and wait times can be long. Determining which incidental findings are worrisome and should be further investigated is key, though little evidence currently exists to guide the radiologist and clinician on further appropriate work up. One classification gaining traction is “interstitial lung abnormalities (ILAs),” which are described as incidental, non-dependent, interstitial abnormalities involving at least 5% of a lung zone in patients without clinical suspicion of ILD [13].If abnormalities are classified as ILAs, certain recommendations exist for appropriate follow up. When ILAs are identified but imaging is incomplete or inconclusive, ordering a dedicated high-resolution chest CT scan is recommended. If ILAs are confirmed, the guidelines recommend assessment for clinically significant disease including a history and physical examination (not necessarily by a pulmonologist) and/or pulmonary function tests. If these confirm evidence of clinically significant disease, referral to a pulmonologist with experience in ILD is recommended. If not, based on the type of radiologic changes (high risk vs. not high risk for progression) there are recommendations for active surveillance with imaging and pulmonary function testing versus expectant management only [13]. A recent prospective cohort study also suggested that patients who have ILAs in addition to family history of pulmonary fibrosis tend to have progressive ILAs that behaves as a subclinical disease, further suggesting the need to delineate follow up strategies for this population [14]. In the recent American Thoracic Society clinical statement regarding the approach to ILA, the authors propose systematic assessment and documentation of ILAs on CT as a foundation for longitudinal follow-up, including closer follow up for patients with first degree family relatives with pulmonary fibrosis and patients with known underlying connective tissue disease [15].

The incidental radiologic findings in this study cannot be classified as ILAs as information about patient symptoms and location and extent of radiologic abnormalities is not available. An interesting follow up to this study might be looking specifically at incidence of appropriate follow up care when a radiologic study reports “ILAs.”

One strength of this study is the large sample size. By querying the EMR, we were able to review over 200,000 CT scan reports. To our knowledge, this is the first report of its type to look at incidental radiologic findings of ILD in different CT scan types and determine incidence of follow up care. This is a first step in quantifying the delay in diagnosing ILD. Previous literature is scarce with the most notable study from 2019 that has included 146 retrospective patients, examining CT interpretation and its effect on timing of ILD patients to be referred to pulmonology [16].

There are several limitations to this retrospective cohort study. First, the radiologic search terms are not specific to ILD, and since the scans were not reviewed by the authors, some non-ILD scans might have been included. For example, ground glass opacities can be due to several different etiologies (including pulmonary edema and infection), and this represented the largest proportion of scans with radiologic findings of ILD, and findings of ground glass opacities should be interpreted with caution in the setting of suspected ILD. For instance, in one retrospective cohort that identified the etiology behind ground glass opacities in 204 patients, 49% of those patients were confirmed to have ILD as the underlying etiology, which was the second most common etiology behind pulmonary edema [17]. Additional information about the reason for ordering a CT scan, such as indication, specialty of the ordering provider, or CT-protocol, could lend additional insight into how likely the specified abnormalities are to represent ILD, however these data were not captured. Secondly, there are other appropriate ways to further investigate possible radiologic findings of ILD that we might not have picked up in our review, such as scan review for accuracy and extent of changes, history and physical examination by a primary care provider, dedicated CT scan of the chest to look for ILD, or referral to services outside of the health care system. Third, as this retrospective cohort study was necessarily constrained to a timeframe decided upon by the authors, we cannot say that any given patient did not have radiologic findings of ILD present on prior CT scans outside of the study timeframe. Along the same lines, we cannot say if any of the included or excluded CT scans in the study timeframe were ordered as follow up scans to abnormalities identified on the index CT (e.g. a dedicated CT chest could have been ordered to better characterize radiographic findings of ILD identified on CT abdomen/pelvis). Finally, we recognize that lowering the threshold to refer incidental ILD findings to pulmonary clinic or pulmonary function testing will surely overwhelm these services (over 8000 patients per year in our cohort derived from 41000 patients over 5 years), and needless to say, not all radiologic abnormalities are clinically meaningful, and some may represent transient, nonspecific, or non-progressive findings. For which reason, further research is needed to develop triage systems for such incidental findings to aid radiologists and CT scan ordering providers in referring patients at higher risk for ILD progression and investigate outcomes for patients with delay in referral to pulmonology after incidental ILD findings. Factoring in electronic consultations with pulmonologists for incidental findings is another potential tool to be investigated in aiding such triage algorithms. One promising avenue is the use of quantitative imaging algorithms. For example, a recent study using lung cancer screening low-dose CT data demonstrated that quantitative measures of interstitial abnormalities (via CALIPER and deep-learning algorithms) were independently associated with all-cause mortality and lung cancer–specific mortality [18]. These findings suggest that quantitative imaging may be able to differentiate high-risk from low-risk individuals, thereby identifying those most likely to benefit from early pulmonary evaluation while reducing unnecessary strain on ILD services.

Conclusions

Incidental radiologic findings of interstitial lung disease on CT scans of the chest, abdomen/pelvis, neck and coronary arteries were common. However, additional work up of the incidental findings was uncommon, with only 16.1% of patients having had pulmonary function testing or pulmonary referral at one year, and 23.9% at 5 years. We aim to educate primary care providers and hospitalists about the importance of evaluating incidental ILD findings from CT scan reports, and identify the need for future research on establishing triage algorithms for such incidental findings to avoid overwhelming pulmonary clinic and PFT laboratories. Improved recognition of incidental ILD findings and referral to the appropriate follow up evaluation are important to decrease diagnostic delay, which can impact ILD morbidity and mortality.

Abbreviations

CT

Computed tomography

ILD

Interstitial lung disease

ILA

Interstitial lung abnormality

PFT

Pulmonary function test

Authors’ contributions

H.K. conceived of the presented idea and supervised the project. A.T. was responsible for querying the EMR and data retrieval. K.J., A.T. and T.M performed initial data analysis. K.T. performed the statistical analysis and created the tables and figures. R.K wrote the manuscript with the assistance of A.K., K.T., and H.K.

Funding

Research reported in this publication was supported by NIH grant P30 CA77598 utilizing the Biostatistics Core shared resource of the Masonic Cancer Center, University of Minnesota and by the National Center for Advancing Translational Sciences of the National Institutes of Health Award Number UM1TR004405. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Data availability

The raw data that support the findings of this study were obtained from the Academic Health Center Information Exchange (AHC IE) of the University of Minnesota, but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. For requesting data, please reach out to [bpic@umn.edu] and the request will be triaged to the corresponding analyst who will process permissions for raw data sharing and access.

Declarations

Ethics approval and consent to participate

This study was approved by the Institutional Review Board of the University of Minnesota. We confirm that all study methods were performed in accordance with the relevant guidelines and regulations of our IRB. Informed consent was waived by the University of Minnesota IRB as no patient health care records were accessed. Research opt-out patients were not included.

Consent to publication

Not applicable.

Competing interests

The authors declare no competing interests.