Impact of the COVID-19 Pandemic on Inpatient Colorectal Cancer Treatment in Japan: An Interrupted Time Series Analysis Using a National Administrative Database

Abstract

Background:

The COVID-19 pandemic disrupted elective endoscopy and non-urgent procedures, but its impact on inpatient colorectal cancer (CRC) treatment in Japan remains incompletely quantified.

Methods:

We conducted an interrupted time series analysis using the Diagnosis Procedure Combination (DPC) inpatient database. Eligible admissions were patients hospitalized between April 2018 and December 2021 with colorectal cancer codes from the International Classification of Diseases, 10th Revision (ICD-10; C18, C19, and C20) who underwent endoscopic submucosal dissection (ESD) or colorectal surgery. We excluded planned chemotherapy readmissions, recurrence cases, admissions with multiple non-ESD surgeries, and records with missing tumor-node-metastasis (TNM) information. Three periods were defined: Period 1 (Apr 2018–Apr 2020), Period 2 (May 2020–Apr 2021), and Period 3 (May 2021–Dec 2021). Monthly inpatient case volumes were modeled using segmented regression with Prais–Winsten estimation and seasonal adjustment.

Results:

Of 276,458 eligible admissions, 260,919 were analyzed. At the start of Period 2, adjusted inpatient cases decreased significantly for ESD (−26%; 887 adjusted vs 1,194 predicted; p<0.001) and colectomy (−6%; 2,746 vs 2,928; p=0.043), whereas rectal resection/abdominoperineal resection showed no significant level change (−5%; p=0.189). In stage-stratified analyses limited to surgical resections, Stage I surgical volume decreased significantly at the start of Period 2 (−13%; p=0.010) with no clear reduction for Stages II–IV.

Conclusion:

In Japan, the early COVID-19 period was associated with short-term reductions in inpatient ESD and Stage I surgical volume, whereas surgical treatment volumes for more advanced disease showed no clear immediate reduction after the first emergency declaration. Because this was an administrative, inpatient-only analysis, the study does not directly assess diagnostic delays, stage migration, or downstream oncological outcomes.

Introduction

The COVID-19 pandemic led to abrupt and widespread disruption of elective medical services, including postponement of non-urgent endoscopy and clinic activity[1-3]. Because gastrointestinal endoscopy can generate aerosols and requires close contact, endoscopy units adopted triage and deferral policies during surges[2,3]. Delays in screening and early diagnosis raised concern for later-stage presentation of colorectal cancer (CRC) and potential downstream increases in CRC-related morbidity and mortality[4,5].

Japan declared a state of emergency in April 2020, which was followed by temporary restrictions of elective endoscopy and non-urgent procedures in many institutions[3,6]. Prior work has suggested that CRC screening participation in Japan fell sharply during the first wave, and international data indicate large temporary reductions in colonoscopy volume during early 2020[3,7]. However, a comprehensive national assessment of inpatient CRC treatment volume—particularly stratified by treatment modality and stage—remains limited.

Using a large national Diagnosis Procedure Combination (DPC) inpatient claims dataset, we aimed to quantify changes in monthly inpatient volumes of CRC treatment (endoscopic submucosal dissection [ESD], colectomy, and rectal resection/abdominoperineal resection) across pandemic phases. We additionally evaluated stage-specific trends to examine whether inpatient treatment volumes changed differently between early-stage and more advanced disease.

Methods

We conducted a retrospective observational study using Japan's Diagnosis Procedure Combination (DPC) database, a large national administrative inpatient dataset that captures hospitalization-level diagnoses and procedures for acute-care settings[8]. We used DPC data because it enables system-level, month-by-month monitoring of inpatient CRC treatments across a large number of hospitals, which is difficult to achieve with single-institution registries during a rapidly evolving public health emergency. Approximately 1,700 hospitals across Japan participate in the DPC system, covering about 85% of acute care beds. This study used all inpatient records for patients admitted between April 2018 and December 2021 from approximately 1,300 hospitals individually collected by the DPC research group.

Eligible admissions were those with colorectal cancer International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) codes C18 (colon), C19 (rectosigmoid junction), or C20 (rectum) and receipt of endoscopic submucosal dissection (ESD) or colorectal surgery during hospitalization. Surgical procedures were identified from the procedure records based on whether the corresponding procedure code was billed during the admission. The primary analytic window was April 2018 through December 2021. Patient selection is summarized in Figure 1.

Figure 1.

Sequential study flow diagram for colorectal cancer (ICD-10 C18–C20) inpatient admissions in the DPC database (Apr 2018–Dec 2021) treated with ESD or colorectal surgery. Records were identified, inclusion criteria applied, and exclusions removed to reach the final analytic cohort (260,919 admissions). Counts for individual exclusion reasons may overlap.

We sequentially applied the following exclusions: planned readmission for chemotherapy, recurrence cases, admissions with multiple non-ESD surgeries, and records with missing tumor-node-metastasis (TNM) components required to derive stage (Figure 1). Some records contained internally inconsistent TNM combinations that prevented reliable stage derivation, such as TisN1M1. These records were categorized as “Unknown” to assess whether changes in stage-classifiable cases were accompanied by major changes in the unknown-stage category.

We defined three analytic periods:

Period 1: Apr 2018–Apr 2020

Period 2: May 2020–Apr 2021

Period 3: May 2021–Dec 2021

Japan's first state of emergency declaration began on April 7, 2020. Considering that the procedures targeted were elective in nature and that institutional restrictions were implemented with some delay, we set the start of Period 2 to May 2020. Period 3 was defined to begin in May 2021, corresponding to the third state of emergency and the subsequent, near-contiguous fourth declaration that expanded to a larger number of prefectures. We note that this boundary is less sharp than the first declaration because regional policies and hospital-level capacity restrictions varied over time. Primary outcomes were monthly inpatient case counts for ESD, colectomy, and rectal resection/abdominoperineal resection. Secondary outcomes were monthly inpatient surgical case counts stratified by clinical stage. ESD cases were analyzed separately and were not included in the stage-stratified surgical analyses. Therefore, the Stage 0 counts in the stage-stratified analysis represent only surgically treated Stage 0 cases and should not be interpreted as the total number of Stage 0 colorectal cancer treatments, because most endoscopically treated early lesions were captured in the ESD category.

Statistical analysis

We performed an interrupted time series analysis using segmented regression to estimate (i) immediate level changes (step changes) and (ii) slope changes (changes in the month-to-month trend) at the transitions between Period 1, Period 2, and Period 3[9,10]. The model included terms for time, indicators for each intervention point, and post-intervention time trends for each period. To account for first-order autocorrelation in the error terms, we fitted a Prais–Winsten regression model with first-order autoregressive [AR(1)] disturbances[11]. Residual autocorrelation was assessed using the Durbin–Watson statistic and residual autocorrelation plots, and was minimal after the AR(1) correction (Durbin–Watson statistic close to 2). To adjust for seasonal variation, we included month indicator variables in the model.

For each of the three procedures (ESD, colectomy, and rectal resection/abdominoperineal resection), we calculated the average monthly number of inpatient cases in each period. We extracted the adjusted value (model-based estimates) for inpatient cases in the first month of Period 2 and Period 3 from the corresponding model. We derived the counterfactual predicted number of inpatient cases at the start of each period transition and quantified the immediate level change and the change in slope between periods. For slope changes, we reported absolute differences in monthly slope coefficients between adjacent periods, rather than percentage changes, because percentage changes can be unstable when the baseline slope is close to zero. For stage-stratified analyses, we applied the same approach to monthly inpatient surgical resection counts stratified by clinical stage: Stage 0, Stage I, Stage II, Stage III, Stage IV, and Unknown. Stata Statistical Software Release 18 (StataCorp LLC, College Station, TX, USA) was used for all statistical analyses. We considered p<0.05 as statistically significant.

Ethics

This study analyzed anonymized secondary administrative data. The study protocol was approved by the ethics committee of University of Occupational and Environmental Health, Japan (Approval No. R4-045).

Results

Among 276,458 eligible admissions, 15,539 were excluded by the prespecified criteria, leaving 260,919 admissions for analysis (Figure 1). Exclusions were applied sequentially in the flow diagram; when reporting counts for each exclusion reason, categories may overlap because a single record could meet more than one exclusion criterion (planned chemotherapy readmission 484; recurrence 12,437; multiple non-ESD surgeries 2,704; missing TNM information 11,086). Baseline characteristics were broadly stable across periods (Table 1). Mean age was approximately 70.3–70.6 years and 56–57% of patients were male. Diagnosis distribution was consistent (C18 67%, C19 5%, C20 28%). The proportion receiving ESD increased modestly over time (18% to 20%) (Table 1).

Table 1.

Baseline Characteristics of Colorectal Cancer (ICD-10 C18–C20) Inpatient Admissions Treated with ESD or Colorectal Surgery in the DPC Database, by Study Period (Period 1–3).

Characteristic	Period 1 (Apr 2018-Apr 2020) N=146,252	Period 2 (May 2020-Apr 2021) N=69,014	Period 3 (May 2021-Dec 2021) N=45,653
Months, n	25	12	8
Age, mean (SD), years	70.3 (11.4)	70.6 (11.5)	70.6 (11.6)
Male, n (%)	83,550 (57%)	39,213 (57%)	25,651 (56%)
Length of stay, mean (SD), days	17.1 (15.1)	16.3 (14.2)	15.7 (12.8)
Primary diagnosis on admission (ICD-10), n (%)
C18: Colon	98,144 (67%)	46,353 (67%)	30,674 (67%)
C19: Rectosigmoid junction	7,673 (5%)	3,605 (5%)	2,338 (5%)
C20: Rectum	40,435 (28%)	19,056 (28%)	12,641 (28%)
Procedure type, n (%)
Endoscopic submucosal dissection (ESD)	26,894 (18%)	13,391 (19%)	9,289 (20%)
Colectomy	77,089 (53%)	35,748 (52%)	23,306 (51%)
Rectal resection/abdominoperineal resection	42,269 (29%)	19,875 (29%)	13,058 (29%)
Cancer stage among surgical resections, n (%)
Stage 0	2,008 (2%)	880 (2%)	548 (2%)
Stage I	30,912 (26%)	13,944 (25%)	8,805 (24%)
Stage II	29,906 (25%)	14,107 (25%)	9,258 (25%)
Stage III	36,780 (31%)	17,171 (31%)	11,802 (32%)
Stage IV	13,275 (11%)	6,415 (12%)	4,033 (11%)
Unknown	6,477 (5%)	3,106 (6%)	1,918 (5%)

Data are shown as mean (SD) or N (%). Percentages for clinical stage were calculated among surgical resection cases only; ESD cases were excluded from the stage distribution and analyzed separately.

Mean length of stay, expressed as mean (standard deviation), decreased across periods: 17.1 (15.1) days in Period 1, 16.3 (14.2) days in Period 2, and 15.7 (12.8) days in Period 3. When stratified by procedure type, mean length of stay for ESD was 6.18 (3.82) days in Period 1, 5.96 (3.24) days in Period 2, and 5.91 (3.25) days in Period 3. For surgical resections, mean length of stay was 19.51 (15.56), 18.76 (14.69), and 18.19 (13.16) days, respectively.

Segmented regression results are summarized in Table 2 and visualized in Figures 2–4. At the start of Period 2, a significant level reduction was observed for ESD (−26%: adjusted 887 vs predicted 1,194; p<0.001) and colectomy (−6%: adjusted 2,746 vs predicted 2,928; p=0.043). Rectal resection/abdominoperineal resection showed a non-significant level change at the same transition (−5%; p=0.189) (Table 2; Figure 4). During Period 2, slope estimates for ESD and colectomy shifted upward relative to Period 1 (Table 2), suggesting a partial rebound (i.e., increasing month-to-month volumes) following the initial decline.

Table 2.

Segmented Interrupted Time Series Results for Monthly Inpatient Volumes of ESD, Colectomy, and Rectal Resection/abdominoperineal Resection.

Procedure	Period	Monthly inpatient Mean (SD)	Predicted inpatient cases at the first month (95% CI)	Adjusted volume for inpatient cases at the first month (95% CI)	Level change	P	Coefficient (95% CI)	Change in slope vs previous period, cases/month	P
Endoscopic submucosal dissection (ESD)	Period 1	1,076 (112)	-	903 (839 to 966)	-	-	11.5 (8.3 to 14.7)	-	-
	Period 2	1,116 (159)	1,194 (1120 to 1269)	887 (840 to 934)	-26%	<0.001	35.8 (27.2 to 44.3)	+24.3	<0.001
	Period 3	1,161 (77)	1,316 (1227 to 1406)	1,123 (1,043 to 1,202)	-15%	0.006	5.6 (-13.2 to 24.5)	-30.2	0.001
Colectomy	Period 1	3,084 (169)	-	3,202 (3,061 to 3,343)	-	-	-10.4 (-18.2 to -2.6)	-	-
	Period 2	2,979 (169)	2,928 (2766 to 3089)	2,746 (2,648 to 2,843)	-6%	0.043	38.9 (16.6 to 61.2)	+49.3	<0.001
	Period 3	2,913 (122)	3,212 (2989 to 3436)	2,902 (2,749 to 3,055)	-10%	0.022	-8.1 (-42.9 to 26.7)	-47.0	0.014
Rectal resection/abdominoperineal resection	Period 1	1,691 (118)	-	1,769 (1,623 to 1,915)	-	-	-4.9 (-11.4 to 1.7)	-	-
	Period 2	1,656 (88)	1,711 (1585 to 1836)	1,624 (1,555 to 1,692)	-5%	0.189	21.6 (6.1 to 37.1)	+26.5	0.001
	Period 3	1,632 (114)	1,883 (1723 to 2043)	1,748 (1,644 to 1,851)	-7%	0.161	-8.0 (-32.1 to 16.2)	-29.6	0.022

Predicted and adjusted values at each period transition are shown with level and slope changes (95% CI, P-values). Models used Prais–Winsten estimation with seasonal adjustment.

Figure 2.

Monthly inpatient ESD volume for colorectal cancer (Apr 2018–Dec 2021). Observed cases and model-predicted trend (with uncertainty band) are shown; dashed lines indicate transitions between Periods 1–3.

Figure 4.

Monthly inpatient rectal resection/abdominoperineal resection volume for colorectal cancer (Apr 2018–Dec 2021). Observed cases and model-predicted trend (with uncertainty band) are shown; dashed lines indicate transitions between Periods 1–3.

Figure 3.

Monthly inpatient colectomy volume for colorectal cancer (Apr 2018–Dec 2021). Observed cases and model-predicted trend (with uncertainty band) are shown; dashed lines indicate transitions between Periods 1–3.

Stage-stratified results are shown in Table 3. At the start of Period 2, Stage I volume decreased significantly (−13%; p=0.010), whereas Stages II–IV showed no clear level decrease at this transition. At the start of Period 3, Stage IV demonstrated a significant level decrease (−16%; p=0.001). Slope estimates in Period 2 suggested increasing Stage IV volumes before Period 3 (Table 3).

Table 3.

Stage-stratified Interrupted Time Series Results for Monthly Inpatient Surgical Volumes (Stage 0, I, II, III, IV, and Unknown).

TNM Stage	Period	Monthly inpatient Mean (SD)	Predicted inpatient cases at the first month (95% CI)	Adjusted volume for inpatient cases at the first month (95% CI)	Level change	P	Coefficient (95% CI)	Change in slope vs previous period, cases/month	P
Stage 0	Period 1	80 (13)	-	102 (89 to 114)	-	-	-0.7 (-1.3 to -0.1)	-	-
	Period 2	73 (12)	68 (55 to 81)	71 (61 to 81)	4%	0.730	-0.1 (-2.0 to 1.8)	+0.6	0.497
	Period 3	69 (6)	70 (53 to 86)	73 (60 to 86)	5%	0.714	-0.8 (-3.8 to 2.3)	-0.7	0.682
Stage I	Period 1	1,237 (84)	-	1,246 (1,181 to 1,311)	-	-	-1.5 (-6.7 to 3.7)	-	-
	Period 2	1,162 (142)	1,253 (1,137 to 1,368)	1,087 (983 to 1,191)	-13%	0.010	19.1 (3.6 to 34.7)	+20.6	0.017
	Period 3	1,101 (69)	1,317 (1,174 to 1,460)	1,185 (1,109 to 1,261)	-10%	0.054	-4.0 (-21.6 to 13.6)	-23.1	0.054
Stage II	Period 1	1,196 (72)	-	1,241 (1,159 to 1,323)	-	-	-4.6 (-9.4 to 0.1)	-	-
	Period 2	1,176 (63)	1,129 (1,029 to 1,228)	1,112 (1,017 to 1,208)	-1%	0.744	9.7 (-3.5 to 22.8)	+14.3	0.043
	Period 3	1,157 (68)	1,228 (1,107 to 1,349)	1,182 (1,078 to 1,287)	-4%	0.524	-11.5 (-36.5 to 13.5)	-21.2	0.121
Stage III	Period 1	1,471 (101)	-	1,577 (1,491 to 1,664)	-	-	-6.1 (-10.0 to -2.2)	-	-
	Period 2	1,431 (91)	1,409 (1,323 to 1,494)	1,354 (1,275 to 1,433)	-4%	0.302	18.5 (5.9 to 31.1)	+24.6	<0.001
	Period 3	1,475 (80)	1,576 (1,459 to 1,693)	1,486 (1,405 to 1,568)	-6%	0.248	-3.5 (-22.7 to 15.7)	-22.0	0.028
Stage IV	Period 1	531 (52)	-	559 (536 to 582)	-	-	-2.0 (-3.4 to -0.6)	-	-
	Period 2	535 (37)	518 (482 to 554)	509 (453 to 566)	-2%	0.728	6.8 (0.7 to 12.8)	+8.8	0.006
	Period 3	504 (37)	591 (547 to 635)	495 (450 to 539)	-16%	0.001	1.2 (-7.3 to 9.6)	-5.6	0.281
Unknown	Period 1	259 (17)	-	254 (232 to 277)	-	-	-0.6 (-1.4 to 0.3)	-	-
	Period 2	259 (28)	249 (222 to 277)	240 (220 to 260)	-4%	0.467	2.8 (-0.3 to 5.9)	+3.4	0.019
	Period 3	240 (15)	274 (238 to 310)	246 (220 to 272)	-10%	0.197	-3.2 (-9.2 to 2.7)	-6.0	0.024

Predicted and adjusted values at each transition are shown with level and slope changes (95% CI, P-values). The change in slope was calculated as the absolute difference in monthly slope coefficients between the current and preceding period and is expressed as cases per month. Prais–winsten estimation with seasonal adjustment was applied. Table 3 includes surgical resections only; ESD cases are excluded and analyzed separately.

Discussion

In this interrupted time series analysis using a national administrative inpatient database, we identified a marked, immediate decline in inpatient ESD volume and a modest decline in colectomy volume at the onset of the early pandemic period (Period 2). Rectal resection/abdominoperineal resection showed a similar direction of change, although the level change did not reach statistical significance. In contrast, inpatient volumes of procedures for more advanced stages showed no clear immediate reduction at the same transition. Clinically, this pattern is compatible with triage and prioritization during resource-constrained surges, whereby deferrable treatment for early-stage lesions may be more susceptible to postponement than treatment for advanced disease. Importantly, because this is an observational analysis of inpatient administrative data, our results quantify changes in treated inpatient case volume rather than directly demonstrating changes in patient prognosis or long-term oncological outcomes.

The procedure-stratified length-of-stay analysis showed modest decreases in both ESD and surgical resections during Period 2. A similar pattern was observed for scheduled admissions in the cardiovascular field[12]. These findings may reflect changes in inpatient care processes during the pandemic, including pressure to secure hospital beds, efforts to reduce in-hospital infection risk, expedited discharge planning, and possible changes in patient selection. However, because the absolute reductions were small and because DPC records are generated at discharge, these results should be interpreted as descriptive indicators of healthcare delivery changes rather than as evidence of clinically substantial reductions in hospitalization burden.

For endoscopists and colorectal specialists, the disproportionate reduction in ESD is particularly informative because ESD is tightly coupled to the screening–diagnosis–curative treatment pathway. When upstream activities are restricted—by institutional policy, resource reallocation, infection-control measures, or patient avoidance—procedural volumes for elective endoscopic cure can drop rapidly[2,3,7].

We observed immediate level decreases in ESD volume and surgically treated Stage I cases at the start of Period 2. In the stage-stratified surgical analysis, Stage I represented surgically treated cases only, whereas early lesions treated by ESD were captured separately in the ESD analysis. Therefore, the reduction in Stage I surgical volume should be interpreted together with the decline in ESD volume as descriptive evidence of reduced inpatient treatment for early-stage or early-lesion disease, rather than as evidence of a specific mechanism such as endoscopic deferral alone.

The slope increases observed during Period 2 likely reflect service adaptation and partial “catch-up,” including restoration of elective capacity through infection-prevention workflows, triage protocols, and backlog clearance[7]. Nevertheless, our observation window ended in December 2021, and the absence of a statistically clear overshoot in Period 3 for ESD or surgery suggests that the follow-up period may be insufficient to fully capture longer-term catch-up of deferred early lesions or downstream changes in case-mix. Longer post-pandemic observation is needed to evaluate whether suppressed early-stage treatment volumes were later compensated and whether any stage migration occurred at diagnosis.

The significant level decrease in Stage IV surgical volume at the start of Period 3 should be interpreted cautiously. The Period 3 boundary was less distinct than the first emergency declaration, regional restrictions varied over time, and discharge-based DPC records may undercount ongoing long-stay admissions near the end of the observation window. Therefore, we regard this Period 3 finding as descriptive rather than conclusive evidence of a specific clinical or policy mechanism.

The Stage 0 estimates also require particularly cautious interpretation. In this study, the stage-stratified analysis was limited to surgical resections and excluded ESD. Therefore, surgically treated Stage 0 cases represented only a small subset of early colorectal lesions and should not be interpreted as the total burden of Stage 0 disease. The small, non-significant positive level changes observed for surgically treated Stage 0 cases in Periods 2 and 3 were based on low monthly counts and do not contradict the reduction in ESD volume or Stage I surgical volume. Accordingly, we did not use Stage 0 trends as primary evidence of pandemic-related disruption.

From a CRC control perspective, the findings underscore the importance of maintaining access to diagnostic colonoscopy and timely definitive treatment for early lesions, even during public health emergencies. National administrative datasets can complement institutional dashboards by enabling system-level monitoring; abrupt declines in ESD and early-stage volumes may serve as early warning signals of upstream pathway disruption, supporting targeted countermeasures to preserve high-value early detection and curative treatment capacity[2,3].

Several limitations should be acknowledged. First, as an observational analysis of administrative data, residual confounding by concurrent secular trends or regional policy heterogeneity is possible, and DPC data do not include outpatient screening/diagnostic endoscopy or detailed clinical variables; therefore, we cannot directly evaluate delays from diagnosis to treatment or long-term oncological outcomes. Second, DPC records are generated at discharge, which may lead to undercounting of ongoing long-stay hospitalizations near the end of the observation window and may contribute to apparent late-period reductions in volume and length-of-stay estimates. Third, although we defined Period 2 and Period 3 based on major state-of-emergency phases, the Period 3 boundary is less distinct because restrictions varied by region and time. Finally, the study ended in December 2021 and may be insufficient to capture longer-term catch-up of deferred early-stage care.

Conclusions

The early COVID-19 pandemic period in Japan was associated with significant short-term reductions in inpatient ESD and surgically treated Stage I colorectal cancer volume, while surgical treatment volumes for Stages II–IV showed no clear immediate reduction after the first emergency declaration. These findings highlight the value of system-level monitoring of inpatient CRC treatment during public health emergencies and the need to preserve essential diagnostic and early-treatment capacity.

Declarations

Funding: This study was funded by a Health, Labour and Welfare Sciences Research Grant from the Ministry of Health, Labour and Welfare of Japan (grant no.: 24AA2006).

Conflicts of Interest

None.

Data Availability

DPC data are not publicly available due to data use agreements; aggregated results may be shared upon reasonable request and approval.

Author Contributions

KN and JN drafted the manuscript. MO collected the data and performed the statistical analyses. HS, AK, YA, MY, KH, YF, KF, and SM critically reviewed the draft and provided important intellectual input. All authors read and approved the final manuscript.

Ethics Approval

The study protocol was approved by the Institutional Review Board of University of Occupational and Environmental Health, Japan (Approval No. R4-045).