Drivers of Variation in the Rate of Radiotherapy Following Lumpectomy in the Military Health System

Abstract

Introduction

Radiation therapy (RT) following lumpectomy has been shown to improve survival. However, little is known about drivers of RT within the Military Health System (MHS). We examined how factors previously shown to affect RT rates after lumpectomy in the civilian population and factors exclusive to service members such as rank, service branch, and MHS catchment area are associated with undergoing RT within 6 months after lumpectomy.

Materials and methods

We conducted a retrospective cohort study using the 2007–2019 MHS Data Repository (MDR) to obtain data from 9681 service members or dependents who had lumpectomy to treat breast cancer and remained in the system at least 6 months after their lumpectomy. We used Cox-Proportional Hazards to model time until their first RT session and examined how demographic and service-related factors affected rates of RT following lumpectomy.

Results

Within the sample, 57.4% of lumpectomy patients received RT within 6 months after surgery. Receiving timely RT was found to be positively associated with older age (Cox Model Benefit Ratio 95% CI [1.01, 1.02]), senior rank (Benefit Ratio 95% CI [1.00, 1.31]), and having a Mental Health diagnosis (Benefit Ratio 95% CI [1.40, 1.60]) when controlling for all other variables in the model. Using catchment area RT rates from 2007 to 2017 and modelling data from 2018 to 2019, it was found that timely RT is positively associated with being in the catchment areas at the top quartile of historical RT rates (Benefit Ratio 95% CI [1.06, 1.47]).

Conclusion

We identify important variation in receiving RT after lumpectomy both when comparing MHS rates to known rates in the general population and when comparing rates across catchment areas where military personnel are stationed. Lumpectomy patients in the MHS who are stationed in catchment areas with higher historical RT rates are themselves more likely to undergo RT.

Introduction

For early-stage breast cancer, substantial evidence suggests that breast conserving surgery (BCS) or lumpectomy followed by radiation therapy (RT) is associated with survival outcomes equivalent to or better than full mastectomy. Full mastectomy has an overall survival hazard ratio of 1.79 and breast cancer-specific survival hazard ratio of 1.66 compared to BCS with RT.^1–3 BCS may be chosen in part to improve psychological and quality of life outcomes resulting from the less disfiguring treatment.^4–6 However, a critical follow-up to BCS is subsequent RT. Numerous studies have demonstrated significant improvement in survival for women undergoing BCS who then receive RT such that it has become the standard guideline for BCS.^2,3,7–12 Yet, despite practice guidelines to provide post-surgical RT (ie adjuvant RT), there is evidence that both individual factors, such as distance,^13,14 as well as practice factors¹⁵ can lead to significant variation in the rates of RT following BCS within the civilian population. However, less is known about RT within the Military Health System (MHS). Given that service members and their dependents typically live where they are stationed, it is important to examine how such assignments impact different facets of their health.

Within the MHS, prior studies have shown improved survival among breast cancer patients within the MHS compared to the civilian setting, in part due to earlier stage of diagnosis.¹⁶ However, evidence suggests lower rates of RT within the MHS.¹⁶ The MHS provides a unique setting to examine drivers of rates of RT following BCS because of universal health insurance coverage within the MHS compared to the civilian sector, where health coverage is a major factor for RT rates in the civilian population.^17,18 We therefore have two objectives in this paper: first, we seek to examine which patient-level factors are associated with increased rates of RT following BCS in the MHS; second, we seek to examine the extent to which service members’ catchment area assignment is associated with RT with the goal of better understanding how to improve guideline concordant breast cancer care within the MHS.

Methods

Data

We conducted a retrospective cohort study using the 2007–2019 MHS Data Repository (MDR) data, which aggregates and validates health care data from the MHS.¹⁹ The MDR includes data on beneficiary characteristics, health care utilization, clinical factors, and enrollment. The utilization information includes claims data on all inpatient and outpatient care received by service members and their enrolled dependents.

We define the analytic sample for all MHS beneficiaries who had a lumpectomy following a breast cancer diagnosis. The different International Classification of Diseases, 9th Revision (ICD-9) and International Classification of Diseases, 10th Revision (ICD-10) diagnosis codes and Current Procedural Terminology (CPT) codes used for identifying the sample were collected from relevant sources^20–28 and enumerated in Appendix Table 1. To ensure we have sufficient observation time to observe RT following lumpectomy, we only included members who remained enrolled in the MHS for at least 6 months after their lumpectomy. Specifically, we wanted to avoid missing RT that may have occurred shortly after diagnosis but was missed because the patient was no longer enrolled in TRICARE coverage so we excluded these observations.

Analysis

We first estimated descriptive measures of the proportion of the cohort that underwent RT within 6 months following lumpectomy across demographic and health measures including age, race/ethnicity (only available for service members), and marital status; service variables such as rank and service branch (either the patient if a service member or the service member spouse if a dependent spouse) and health variables, particularly any mental health (MH) diagnosis, and if the patient underwent chemotherapy or endocrine therapy as part of treatment for breast cancer. Next, we estimated univariate Kaplan–Meier (KM) time-to-event curves for each of the demographic and health variables to describe differences in the trends that exist in the cohort with respect to time between lumpectomy and the initiation of RT. The demographic and health variables were selected as confounders based on previous studies that identified these as factors that affected patients’ undergoing of RT after lumpectomy.^29–32

Third, we estimated a Cox proportional hazards (Cox-PH) model to estimate the association of the demographic and health variables on months to RT since lumpectomy. This model was preferred since we are interested in both the occurrence of RT and its timeliness as per standard of care. Since the outcome is desirable (earlier RT), we labeled the cox hazard rate as a benefit rate. We then estimated a similar model that focused on RT between 2018 and 2019 so that we could add a measure of catchment area RT. Specifically, we included quartile of 2007–2017 catchment area RT as a lagged measure of RT meant to capture whether a patient living in a catchment area with a higher historical rate of RT was more likely to receive RT in subsequent years. To further understand how catchment area impacted RT rates, we plotted catchment area lumpectomy and RT rates along with the correlation between number of lumpectomies and RT rate. Further, we plotted 2018–2019 RT rates by catchment area quartile of RT rate (using 2007–2017 rates). We then divided catchment areas into the top and bottom quartile of 2007–2017 to plot a Kaplan–Meier curve that describes the difference in RT experience for service members belonging to each catchment area grouping who had a lumpectomy in 2018–2019.

All analyses were conducted using a 0.05 significance level using R 4.5.1 and the survival and survminer packages. Our study complies with the Declaration of Helsinki and was approved by the Penn State Institutional Review Board (IRB) (Study #00014029) and by the Defense Health Agency Office of Research Protections Human Research Protection Official (DHQ-20-2044). Data access protocols used comply with relevant data protection and privacy regulations. Given the retrospective nature of the study and that only de-identified data were received, the IRB waived the informed consent requirement.

Results

Of all patients within the sample who had a breast cancer diagnosis from 2007 to 2019 (n=55,533), there were 14,234 patients who underwent lumpectomy. Among them, there were 9681 who remained in the MHS at least 6 months after their lumpectomy. Among these, 5554 (57.4%) had at least one RT session within 6 months after their lumpectomy (see Appendix Figure 1).

Table 1 compares the characteristics of women who received RT following a lumpectomy to those who did not receive RT. The mean age in years at lumpectomy was 48 (sd=9). We found that women receiving RT were significantly older (Table 1A), more likely a senior officer (or married to one), retired (Table 1B), had a mental health diagnosis, or had undergone chemotherapy or endocrine therapy (Table 1C). We did not observe differences in race/ethnicity (Table 1A).

Table 1.

Rate of Radiotherapy Following Lumpectomy

A. Demographic Factors
Race/Ethnicity	RT		No RT		Chi-sq p-value
	n	Percent	n	Percent
Non-Hispanic White	875	16%	662	16%	0.96
Non-Hispanic Black	434	8%	330	8%
Hispanic	36	1%	26	1%
Other/Unknown	4208	76%	3110	75%
Age
<40	838	15%	937	23%	<0.0001
40–49	2014	36%	1482	36%
50–59	2140	39%	1369	33%
≥60	561	10%	335	8%
Married
No	3272	79%	4502	81%	0.03
Yes	855	21%	1052	19%
B. Military Service Factors
Rank
Junior enlisted	104	2%	123	3%	<0.0001
Senior enlisted	3964	71%	2966	72%
Junior officer	136	2%	131	3%
Senior officer	1163	21%	756	18%
Warrant Officer	190	3%	143	3%
Status
Active	686	12%	692	17%	<0.0001
Retired	4871	88%	3427	83%
Service branch
Army	2057	37%	1520	37%	0.03
Coast Guard	108	2%	112	3%
Air Force	1812	33%	1307	32%
Marine	318	6%	279	7%
Navy	1223	22%	868	21%
Other	38	1%	34	1%
C. Medical Factors
Mental Health Diagnosis (prior to or following lumpectomy)
No	4368	79%	3559	86%	<0.0001
Yes	1187	21%	567	14%
Chemotherapy (prior to or following lumpectomy)
No	3270	59%	2675	65%	<0.0001
Yes	2283	41%	1453	35%
Endocrine therapy (following lumpectomy)
No	5498	99%	4063	98%	0.03
Yes	57	1%	63	2%

We found that officers and more senior service members (or spouses) tended to have higher RT rates than enlisted or more junior service members or spouses; retirees or spouses of retirees had higher RT rates than active duty members; and older patients have higher RT rates than younger patients. Among different service branches, the Coast Guard has the lowest RT rates. Members with a mental health diagnosis prior to or after lumpectomy and those undergoing chemotherapy had higher RT rates. Members who had an endocrine therapy session after their lumpectomy had a lower rate of RT, which is consistent with the practice found in literature of endocrine therapy being used instead of RT.^33,34

Figure 1 shows the univariate KM curves which reflect similar trends as seen in Table 1 which shows univariate analysis for undertaking RT for each demographic, health, and service variable. The first panel shows the time-to-event curve for the sample cohort, showing that about 40% of the cohort has their first RT session within 3 months after lumpectomy and about 57% after 6 months. Subsequent panels show comparisons of time to RT across different age groups, rank, sponsor, marital status, chemotherapy treatment, endocrine therapy treatment, and mental health diagnosis. Likelihood ratio tests conducted for each confounding variables showed consistent results with univariate tests in Table 1.

Figure 1.

Univariate Kaplan–Meier curves for months until RT.

Table 2 shows results from the Cox PH model of the time to first RT session for the cohort within 6 months of observation following their lumpectomy. The model included age, race, rank, retirement status, and service branch at the time of lumpectomy, marital status and mental health diagnosis prior and up to 6 months following lumpectomy, and endocrine therapy following lumpectomy. Catchment area was included as a stratification variable. Results show that older age (p<0.0001), senior rank (p<0.05) and having a mental health diagnosis (p<0.0001) were positively associated with time to RT within 6 months after lumpectomy.

Table 2.

Cox Proportional Hazard (Benefit) Model Results of Likelihood of Radiotherapy Following Lumpectomy by Patient Characteristics

	Coefficient	Benefit Ratio	95% CI Lower Bound	95% CI Upper Bound	p-value
Age at lumpectomy	0.02	1.02	1.01	1.02	<0.0001
Race/ethnicity
(Hispanic)	0.21	1.24	0.88	1.74	0.2154
(Non-Hispanic Black)	−0.02	0.98	0.87	1.10	0.7216
(Other/Unknown)	0.00	1.00	0.92	1.08	0.9483
Enlisted	−0.06	0.94	0.88	1.00	0.0674
Senior	0.14	1.15	1.00	1.31	0.0481
Warrant Officer	0.09	1.10	0.90	1.34	0.3557
Retired	0.08	1.09	1.00	1.19	0.0632
Coast Guard	−0.16	0.85	0.70	1.04	0.1091
Air Force	−0.03	0.97	0.91	1.04	0.4411
Marine	−0.03	0.97	0.85	1.10	0.5954
Navy	0.00	1.00	0.92	1.08	0.9276
Other	−0.17	0.84	0.61	1.17	0.3131
Married	−0.08	0.93	0.85	1.01	0.07
Mental Health diagnosis	0.40	1.50	1.40	1.60	<0.0001
Had Endocrine Therapy	−0.20	0.81	0.63	1.06	0.1297
Had Chemotherapy	0.01	1.01	0.96	1.07	0.7147

Notes: Omitted categories include non-Hispanic White race/ethnicity, officer, junior rank, Army, not married.

Figure 2 provides a detailed description of the distributions of lumpectomies and RT rates. The first panel shows that catchment areas most commonly performed between 20 to over 100 lumpectomies across the 12-year period and that RT rates after lumpectomy commonly ranged from 40% to 55%. The third panel shows a positive correlation between the number of lumpectomies and RT rate (r=0.26).

Figure 2.

Catchment area rates of lumpectomy and subsequent radiotherapy.

Next, we examined whether patients living in catchment areas with higher RT rates ended up being more likely to receive RT after lumpectomy. The catchment area RT rates were computed using data from all lumpectomies that occurred between 2007 and 2017 (n=8443). We plotted 2007–2018 catchment area RT rates by quartile in Figure 3. Figure 4 shows that patients who had lumpectomies in catchment areas with historically higher RT rates had a greater chance of receiving RT than patients who had lumpectomies in catchment areas with historically lower RT rates. We find similar results with controlling for all the aforementioned confounders (see Table 3).

Figure 3.

Radiotherapy rate by quartile of catchment area radiotherapy rate.

Figure 4.

Kaplan-Meier curve for months until RT in 2018–2019 by historical (2007–2017) catchment area radiotherapy rate.

Table 3.

Cox Proportional Hazard (Benefit) Model Results of Likelihood of Radiotherapy Following Lumpectomy (2018–2019) by Patient Characteristics and 2007–2017 Catchment Area Rates

	Coefficient	Benefit Ratio	95% CI Lower Bound	95% CI Upper Bound	p-value
Catchment area quartile of RT (2007–2017) Quartile
(1)	0.07	1.08	0.76	1.53	0.6809
(2)	−0.23	0.79	0.66	0.96	0.0176
(3)	−0.22	0.80	0.68	0.94	0.0053
Age at 1st lumpectomy	0.01	1.01	1.00	1.02	0.0566
Race/ethnicity
(Hispanic)	0.71	2.04	1.13	3.69	0.0182
(Non-Hispanic Black)	−0.05	0.95	0.72	1.26	0.7340
(Other/Unknown)	0.04	1.04	0.87	1.26	0.6452
Enlisted	0.16	1.18	1.00	1.39	0.0567
Senior	−0.03	0.97	0.72	1.30	0.8259
Warrant Officer	−0.19	0.82	0.51	1.33	0.4277
Retired	0.21	1.23	0.98	1.54	0.0775
Service branch
(Coast Guard)	−0.21	0.81	0.49	1.32	0.3988
(Airforce)	−0.02	0.98	0.83	1.16	0.8006
(Marines)	0.01	1.01	0.76	1.34	0.9679
(Navy)	0.04	1.05	0.87	1.25	0.6256
(Other)	−0.52	0.60	0.22	1.63	0.3124
Married	−0.06	0.94	0.77	1.14	0.5205
Mental Health diagnosis	−0.02	0.98	0.86	1.13	0.8072
Had Endocrine Therapy	0.92	2.51	0.35	17.97	0.3601
Had Chemotherapy	0.14	1.16	1.01	1.32	0.0378

Notes: Omitted categories include quartile 4 of catchment area RT rates, non-Hispanic White race/ethnicity, officer, junior rank, Army, not married.

Discussion

We found that 57.4% of lumpectomy patients in the MHS from 2007 to 2019 received RT within 6 months after lumpectomy. Adjusting for age, this proportion is lower than those found in studies in the civilian population, which ranges from 60% to more than 90%.^31,35 Looking at demographic and health variables, we found older patients, married patients, and those undergoing chemotherapy and not undergoing endocrine therapy were more likely to undergo RT. Lower income has also been associated with RT refusal in prior studies³¹, and while we did not have income in our data, service variables on rank showed that more senior ranked service members also had a higher likelihood of undergoing RT.

In addition, we found that provider or place characteristics appear to be important drivers of variation in RT rates. Specifically, patients who had lumpectomy in 2018–2019 in a catchment area within the top quartile of RT rates (based on data from 2007–2017) were 20% more likely to receive RT within 6 months than those in lower quartiles at any time during that post-lumpectomy period. Understanding what factors drive this variation, such as specific provider practice styles, hospital procedures, or other system-level factors, is important for learning from catchment areas with high rates of RT in order to improve rates in areas with lower rates of RT.

Our study has several important strengths to note. We have a unique perspective to observe RT utilization with a very large sample of all claims data within the MHS 2007–2019; with which we can observe actual RT. Finally, our study has several important limitations to note. First, much of our results assess associations between factors and rates of RT following lumpectomy and so have limited causal interpretations. However, for the effect of historical catchment area RT rates on subsequent patient RT rates, we note that patients are largely located randomly due to the random nature of stations resulting from a permanent change of station (PCS). Second, we cannot observe the stage of diagnosis or tumor size, a common limitation for claims, data, nor can we observe the reasons for why patients may opt not to undergo RT. Third, racial and ethnic data are missing for dependents so our results by race/ethnicity are limited to service members. Fourth, we may not fully observe dependents’ utilization if they receive care using exclusively another type of insurance, although we note that we focus our analyses on only those with continuous TRICARE coverage. Fifth, while the overall study includes a large number of patients, some of the subsets are relatively small. Specifically, only 36 patients identified as Hispanic and only 57 had endocrine therapy.

Conclusion

Variability of undertaking radiotherapy after lumpectomy among service members is driven by factors similar to those of the general US population. However, the rate at which service members undergo this important standard of care after BCS is lower than the rate in the general population, particularly in cases where service members and their dependents are stationed in catchment areas with historically low rates. This implies a need to understand what modifiable factors drive catchment areas with higher rates of RT to achieve these higher rates of RT in order improve RT rates following lumpectomy in the MHS.

Funding Statement

Defense Health Agency Grant #HQ00342010002.

Data Sharing Statement

Data are not publicly available due to the terms of the data sharing agreement.

Ethical Approval Statement

The study was approved by the Penn State Institutional Review Board (IRB) (Study #00014029) and by the Defense Health Agency Office of Research Protections Human Research Protection Official (DHQ-20-2044).

Informed Consent

Since the study was retrospective and all data received were already de-identified, exemption was granted by the Penn State Institutional Review Board (IRB).

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agreed to be accountable for all aspects of the work.

Disclosure

The authors report no conflicts of interest in this work.