Sex- and age- specific lung cancer incidence and mortality among Medicaid beneficiaries with and without HIV, 2001–2015

Abstract

Background

People with HIV (PWH) have higher lung cancer incidence and poorer prognosis than people without HIV (PWoH). PWoH experience age and sex-based differences in lung cancer. However, evidence on sex- and age-specific trends among PWH are limited.

Methods

We evaluated lung cancer incidence for Medicaid beneficiaries aged 30–64 from 14 states (2001–2015) by HIV status. We used Cox models and inverse probability weighting to estimate the hazard ratio (HR) of lung cancer and the Aalen-Johansen estimator to calculate the cumulative incidence of lung cancer by HIV status. We estimated the excess mortality ratio (EMR) by comparing the ratio of mortality rate differences before and after lung cancer diagnosis by HIV status.

Results

There were 1,110 lung cancers over 529,150 person-years among PWH and 8,256 lung cancers over 8,110,847 person-years among PWoH. HIV was associated with the hazard of lung cancer incidence (HR: 1.75, 95% CI:1.56, 1.96); the strongest association was among female beneficiaries aged 30–49 (HR: 3.71, 95% CI: 2.59, 5.32). Lung cancer incidence rates declined over time for all beneficiaries; however, the smallest decrease was in female PWH. PWH experienced significant excess mortality after lung cancer diagnosis compared to PWoH (EMR:1.24, 95% CI:1.06, 1.43); this was strongest among female beneficiaries aged 30–49 (EMR: 2.18, 95% CI: 1.46, 3.12).

Conclusions

PWH had elevated incidence of lung cancer and excess mortality after lung cancer compared to PWoH. Importantly, these findings were strongest among young female beneficiaries with HIV, highlighting the need for lung cancer prevention and early detection among this group.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12885-026-16487-0.

Background

Lung cancer has emerged as the second most common cancer and the second leading cause of cancer death among people with HIV (PWH) in the United States (US) [1, 2]. Compared to people without HIV (PWoH), PWH have an elevated incidence of lung cancer and tend to be diagnosed at a younger age with more advanced stage [3–9]. Lung cancer is expected to remain one of the most common cancers among PWH, driven by the projected rise in the population of older PWH [10]. This is particularly concerning given that lung cancer has high mortality rates in the general population, and PWH tend to experience worse survival outcomes than PWoH [7, 11–15].

In the general US population, lung cancer rates have historically been higher in men, but the past few decades have seen a steeper decline among men than women [16]. Recent evidence suggests younger women have higher rates than their male counterparts [17–20]. Among people who heavily smoked, one study reported that women still exhibit a higher incidence of lung cancer after adjusting for smoking and age [21]. Changes in smoking-related behaviors in both sexes over time may explain some, but not all, of this phenomenon [18]. However, whether these patterns are consistent among PWH remains unclear.

While the overall positive association between HIV and lung cancer has been well documented, data examining potential differences by sex and age are limited. A recent study in PWH estimated that, after age adjustment, people who inject drugs of both sexes had higher risk of lung cancer, while men who have sex with men and women with other/unknown risk behaviors had lower risk, compared to men with other/unknown risk behaviors [22]. Another study reported greater standardized incidence ratios of lung cancer in younger age groups among PWH [23]. We previously found that that the crude cumulative incidence of lung cancer was higher among PWH than among PWoH enrolled in Medicaid [24]. Notably, this difference by HIV status was consistent for female beneficiaries across all ages, but attenuated for males at older ages [24].

Medicaid is a state-administered program that provides coverage for low-income individuals and those with disabilities in the US. It is also the largest payer for HIV treatment and provides coverage to approximately 40% of US PWH [25]. Medicaid beneficiaries, like PWH [26, 27], have higher smoking rates than the general population [28, 29]. In contrast to comparisons to the general population, comparisons among Medicaid beneficiaries also allow for more comparability in socioeconomic status, access to medical services, and some lifestyle factors by HIV status.

To build upon our prior findings, we more comprehensively examined lung cancer and weighted excess mortality associated with lung cancer, stratified by sex and age among Medicaid beneficiaries with and without HIV enrolled in 14 US states during 2001–2015. We hypothesized that PWH would have a consistently higher burden of lung cancer over time and higher excess mortality after with lung cancer compared to PWoH. We anticipated that the observed disparities in lung cancer incidence (based on crude cumulative incidence) by HIV status, sex, and age in our prior study would persist – although the magnitude may be attenuated when using weighting to better control for confounding and informative right censoring.

Methods

Study sample

We used Medicaid Analytic eXtract (MAX) and Transformed Medicaid Statistical Information System Analytic File (TAF) data from the Centers for Medicare and Medicaid Services (CMS) [30] for Medicaid beneficiaries who enrolled between 2001 and 2015 in 14 US states: Alabama, California, Colorado, Florida, Georgia, Illinois, Massachusetts, Maryland, North Carolina, New York, Ohio, Pennsylvania, Texas, and Washington. These 14 states were chosen to represent the geographical distribution of the North American AIDS Cohort Collaboration on Research and Design (NA-ACCORD) [31].

We restricted to beneficiaries aged 30–64 years with full Medicaid benefits who had ≥ 7 months of continuous eligible enrollment. Eligible enrollment was defined as ≥ 15 days of Medicaid coverage during a month and no dual enrollment in Medicare or private insurance, in order to ensure all healthcare encounters were captured in Medicaid data [24, 32–34]. We required ≥ 7 months of continuous enrollment to allow for a 6-month “washout period” for identification of baseline HIV status and prevalent cancers, as well as ≥ 1 month of follow-up [33, 35]. Additional information on Medicaid and the analytic approach is provided in the Appendix (Additional file 1). The Johns Hopkins Bloomberg School of Public Health Institutional Review Board determined that this study met the criteria for exemption.

We defined analytic baseline as 6 months after the start of beneficiaries’ first continuous enrollment period [32, 33, 35–37] or their 30th birthday, whichever occurred later. All conditions were identified based on relevant ICD-9 or ICD-10 codes recorded in the diagnosis fields of the MAX/TAF Inpatient Files, Other Services Files, and Long Term Care Files. A complete list of conditions and their corresponding ICD codes are provided in Table S1 (Additional File 2). We excluded beneficiaries with prevalent cancers using ICD codes representing any cancer (Table S2, Additional file 2) on an inpatient or outpatient/long-term care claim prior to baseline to enhance capture of first primary cancer diagnoses [24, 32].

Exposure assessment

Our exposure was HIV status at baseline, based on the presence of 1 inpatient claim or 2 outpatient/long-term care claims within two years with an HIV ICD-9 or ICD-10 diagnosis code (Table S1, Additional file 2) [24, 37–39]; the service start date for the first claim was used as the date of HIV diagnosis, meaning a beneficiary was classified as having HIV if the date of HIV diagnosis was before baseline. For computational efficiency, we included all beneficiaries with HIV and a 25% random sample of beneficiaries without HIV from each state [32].

Outcome assessment

Our outcome was the first lung cancer diagnosis. We only considered the first cancer because people’s risk profiles, healthcare interactions, and cancer screening patterns may change after developing the initial cancer. Following the standard CMS Chronic Conditions Warehouse definition [40], we defined incident lung and other cancer diagnoses as having the relevant ICD-9 and ICD-10 codes on 1 inpatient or 2 outpatient/long-term care claims within two years; the service start date for the first claim was used as the date of diagnosis (Table S1, Additional file 2).

Covariate assessment

We obtained race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, or other/unknown), sex (male or female), US state, calendar year of enrollment, and dates of birth and death from the MAX Personal Summary File and the TAF Demographic and Eligibility File. We defined Charlson comorbidities [41] (excluding HIV and cancer) as any relevant ICD-9 or ICD-10 code (Table S1, Additional file 2) on an inpatient or outpatient/long-term care claim; the number of comorbidities (0, 1, or ≥ 2) were time-updated from baseline as a surrogate for overall healthcare needs.

Statistical analysis

Crude incidence rates of lung cancer

Beneficiaries were followed from baseline until of the first lung cancer diagnosis, first diagnosis of any other cancer, death, Medicaid disenrollment, reaching age 65, or 12/31/2015, whichever came first. We estimated lung cancer incidence rates among beneficiaries with and without HIV as the number of cases per 10,000 person-years to describe patterns of the unadjusted real-world burden over time for practical public health relevance. Confidence intervals (CIs) of the crude rates were estimated based on the mid-P exact test [42], as we did not want to enforce the normality assumption, particularly since lung cancer was a rare outcome and event counts were small in certain subgroup analyses. For analyses stratified by age (< 50 or ≥ 50) or calendar period (2001–2005, 2006–2010, or 2011–2015), beneficiaries began to contribute time at risk from the first day after baseline in which they were within the age range or calendar period of the specific category. We chose age 50 as the cut-off based on our prior work showing substantially increased lung cancer incidence above this age in our Medicaid sample [24]. Five-year intervals were used conventionally to examine unadjusted calendar trends. Beneficiaries diagnosed with lung cancer at an earlier age/time period than the minimum age/time period of the category were excluded.

Hazards and cumulative incidence of lung cancer

We used Cox proportional hazards regression to estimate the weighted hazard ratio (HR) of lung cancer incidence by HIV status with age as the time scale; 95% CIs were based on the robust variance estimator. We assessed the proportional hazard assumption statistically with the Schoenfeld Residuals Test and graphically with log(-log(S(t))) plots and Kaplan-Meier curves. There was evidence of quantitative but not qualitative violations of the assumption (i.e., the log(-log(S(t))) curves were not parallel, but the Kaplan-Meier curves did not cross); thus, we interpreted the estimated HRs as time-averaged estimates. We estimated the crude and weighted cumulative incidence functions (CIFs) for lung cancer by HIV status, using the Aalen-Johansen estimator with age as the time scale and treating death as a competing event. We compared the CIFs to estimate risk differences (RDs) and risk ratios (RRs) by HIV status. The 95% CIs were based on the 2.5th and 97.5th percentiles of CIF point estimates from 500 bootstrap resamples for the weighted CIF, and on the extension of Greenwood’s formula for the crude CIF.

When estimating the CIF and HRs, we used stabilized inverse probability weights to control for confounding and informative right censoring conditional on age, sex, race/ethnicity, state, calendar period, and comorbidity; follow-up time was discretized into three-month intervals. Censoring weights also included time-updated comorbidity status; exposure weights were constructed without (weight 1) and with (weight 2) baseline comorbidity status (additional details in the appendix, Additional File 1)).

Because we were unable to adjust for smoking status, we calculated E-values for the HR and RR (the E-value for the RD will match that for the RR, since the E-value is determined by the confounding strength on the RR scale), to quantify the magnitude of an association between a confounder that would be required for the 95% CIs to include the null [43].

Excess mortality associated with lung cancer

To calculate excess mortality by HIV status, we compared the background mortality rate of all-cause mortality among beneficiaries who had not yet developed lung cancer and the all-cause mortality rate after lung cancer diagnosis. For the background mortality rate, we included person-time from all individuals from baseline until the lung cancer diagnosis, death, or censoring, and death counts contributed by only those who died without developing lung cancer. For post-lung cancer diagnosis mortality rates, we only included person-time from those with lung cancer, from diagnosis until death or censoring, and deaths among those with lung cancer. We estimated these rates for PWH and PWoH separately in a weighted pseudopopulation (standardized to the baseline distribution of PWH in our study sample from the 14 states, using exposure odds weights [44]; additional details in the appendix, Additional File 1). We then compared the mortality rate difference (post-diagnosis rate minus background rate) between PWH and PWoH. A ratio > 1 indicates that PWH experience greater excess mortality after lung cancer diagnosis compared to PWoH. The 95% CIs of the weighted mortality rate differences and the weighted ratio were based on 500 bootstrap resamples.

For age or year-stratified analyses, person-time was partitioned based on the time-updated age and calendar period, respectively, and we assessed event status within each time interval. We conducted all analyses overall and stratified by age, sex, and race/ethnicity. Incidence rates were also stratified by calendar period. All analyses were conducted using R version 4.3.1 (R Foundation, Vienna, Austria, 2023-07-19).

Results

We included 162,754 beneficiaries with HIV and 3,740,358 beneficiaries without HIV with a median follow-up of 18 (IQR: 7–50) and 13 (IQR: 6–28) months, respectively. The median age at baseline was similar for PWH and PWoH (Table 1). Compared to PWoH, PWH were more likely to be male, non-Hispanic Black, and have ≥ 1 comorbid conditions at baseline (Table 1).

Table 1.

Baseline characteristics of Medicaid beneficiaries by HIV status

	Beneficiaries without HIV	Beneficiaries with HIV
N	3,740,358	162,754
Age at enrollment (median [IQR])	43.49 [35.11, 53.84]	43.98 [37.39, 51.04]
30–39 (%)	1,507,122 (40.3)	56,329 (34.6)
40–49 (%)	964,265 (25.8)	60,431 (37.1)
50–59 (%)	820,586 (21.9)	32,511 (20.0)
60–64 (%)	448,385 (12.0)	13,483 (8.3)
Sex (%)
Female	2,124,263 (56.8)	56,472 (34.7)
Male	1,616,095 (43.2)	106,282 (65.3)
Race/ethnicity (%)
Non-Hispanic White	1,591,265 (42.5)	40,548 (24.9)
Non-Hispanic Black	763,633 (20.4)	77,288 (47.5)
Hispanic	569,015 (15.2)	15,169 (9.3)
Other/Unknown	816,445 (21.8)	29,749 (18.3)
State (%)
AL	48,113 (1.3)	1733 (1.1)
CA	1,061,539 (28.4)	27,367 (16.8)
CO	52,560 (1.4)	728 (0.4)
FL	236,068 (6.3)	18,869 (11.6)
GA	113,412 (3.0)	6913 (4.2)
IL	307,510 (8.2)	8765 (5.4)
MA	192,596 (5.1)	6972 (4.3)
MD	100,721 (2.7)	6778 (4.2)
NC	115,616 (3.1)	6321 (3.9)
NY	634,574 (17.0)	57,524 (35.3)
OH	240,772 (6.4)	4459 (2.7)
PA	222,332 (5.9)	4556 (2.8)
TX	214,436 (5.7)	8107 (5.0)
WA	200,109 (5.3)	3662 (2.3)
Enrollment period (%)
2001–2005	1,428,263 (38.2)	83,165 (51.1)
2006–2010	703,462 (18.8)	29,432 (18.1)
2011–2015	1,608,633 (43.0)	50,157 (30.8)
Charlson comorbidity count (%)
0	2,934,464 (78.5)	103,157 (63.4)
1	547,998 (14.7)	39,897 (24.5)
2+	257,896 (6.9)	19,700 (12.1)

Lung cancer incidence and mortality: all ages

During follow-up, we observed 1110 lung cancer cases among PWH and 8526 cases among PWoH (Table 2). The crude incidence rates and weighted hazard of lung cancer were substantially higher among PWH than PWoH overall and when stratified by sex, race/ethnicity, and age (Table 2).

Table 2.

Association between HIV and lung cancer by age, sex, and race/ethnicity among Medicaid beneficiaries from 14 states, 2001–2015

	Cases	Person-years	Crude IRa (95% CI)	Weighted HRb (95% CI)	E-value for HR (point estimate/CI)	Weighted HRc (95% CI)	E-value for HR (point estimate/CI)
All ages
Overall
No HIV	8256	8,110,847	10.18 (9.96, 10.4)	Reference		Reference
HIV	1110	529,150	20.98 (19.77, 22.24)	1.85 (1.66, 2.06)	3.1/2.71	1.75 (1.56, 1.96)	2.9/2.49
Female
No HIV	4004	4,923,815	8.13 (7.88, 8.39)	Reference		Reference
HIV	452	220,192	20.53 (18.7, 22.49)	2.20 (1.83, 2.66)	3.82/3.06	1.99 (1.60, 2.47)	3.39/2.58
Male
No HIV	4252	3,187,032	13.34 (12.95, 13.75)	Reference		Reference
HIV	658	308,957	21.3 (19.72, 22.97)	1.52 (1.33, 1.73)	2.41/1.99	1.48 (1.29, 1.70)	2.32/1.9
Non-Hispanic White
No HIV	4507	3,422,687	13.17 (12.79, 13.56)	Reference		Reference
HIV	255	116,787	21.83 (19.28, 24.64)	1.50 (1.25, 1.81)	2.37/1.81	1.44 (1.18, 1.75)	2.24/1.64
Non-Hispanic Black
No HIV	2262	1,957,851	11.55 (11.08, 12.04)	Reference		Reference
HIV	656	262,097	25.03 (23.17, 27)	2.03 (1.81, 2.28)	3.48/3.02	1.93 (1.72, 2.17)	3.27/2.83
Hispanic
No HIV	398	1,189,051	3.35 (3.03, 3.69)	Reference		Reference
HIV	56	47,492	11.79 (8.99, 15.2)	1.97 (1.24, 3.12)	3.35/1.79	1.59 (1.03, 2.46)	2.56/1.21
Non-Hispanic White Female
No HIV	2213	2,021,333	10.95 (10.5, 11.41)	Reference		Reference
HIV	85	37,974	22.38 (17.99, 27.54)	1.82 (1.29, 2.56)	3.04/1.9	1.72 (1.13, 2.62)	2.83/1.51
Non-Hispanic Black Female
No HIV	1058	1,238,047	8.55 (8.04, 9.07)	Reference		Reference
HIV	300	122,159	24.56 (21.9, 27.46)	2.71 (2.30, 3.20)	4.86/4.03	2.48 (2.09, 2.95)	4.4/3.6
Hispanic Female
No HIV	201	750,562	2.68 (2.33, 3.07)	Reference		Reference
HIV	16	17,334	9.23 (5.46, 14.67)	1.67 (0.79, 3.55)	2.73/1	0.95 (0.47, 1.90)	1.29/1
Non-Hispanic White Male
No HIV	2294	1,401,354	16.37 (15.71, 17.05)	Reference		Reference
HIV	170	78,814	21.57 (18.51, 25)	1.33 (1.07, 1.65)	1.99/1.34	1.31 (1.04, 1.64)	1.95/1.24
Non-Hispanic Black Male
No HIV	1204	719,804	16.73 (15.8, 17.69)	Reference		Reference
HIV	356	139,938	25.44 (22.9, 28.19)	1.44 (1.23, 1.69)	2.24/1.76	1.43 (1.22, 1.67)	2.21/1.74
Hispanic Male
No HIV	197	438,489	4.49 (3.9, 5.15)	Reference		Reference
HIV	40	30,158	13.26 (9.6, 17.88)	1.86 (1.16, 2.99)	3.12/1.59	1.71 (1.04, 2.82)	2.81/1.24
Age 30–49 Years
Overall
No HIV	1233	4,878,616	2.53 (2.39, 2.67)	Reference		Reference
HIV	390	331,446	11.77 (10.64, 12.98)	3.32 (2.76, 3.99)	6.1/4.96	3.16 (2.58, 3.88)	5.77/4.6
Female
No HIV	715	3,062,783	2.33 (2.17, 2.51)	Reference		Reference
HIV	199	147,116	13.53 (11.74, 15.51)	3.78 (2.76, 5.18)	7.02/4.96	3.71 (2.59, 5.32)	6.88/4.62
Male
No HIV	518	1,815,832	2.85 (2.61, 3.11)	Reference		Reference
HIV	191	184,330	10.36 (8.97, 11.91)	2.71 (2.11, 3.47)	4.86/3.64	2.54 (1.95, 3.31)	4.52/3.31
Non-Hispanic White
No HIV	653	2,060,591	3.17 (2.93, 3.42)	Reference		Reference
HIV	84	74,573	11.26 (9.04, 13.88)	3.15 (2.26, 4.39)	5.75/3.95	3.00 (2.12, 4.23)	5.45/3.66
Non-Hispanic Black
No HIV	352	1,206,140	2.92 (2.63, 3.24)	Reference		Reference
HIV	242	162,766	14.87 (13.08, 16.83)	3.71 (3.07, 4.48)	6.88/5.59	3.47 (2.85, 4.21)	6.4/5.15
Hispanic
No HIV	80	759,061	1.05 (0.84, 1.3)	Reference		Reference
HIV	16	30,668	5.22 (3.09, 8.29)	1.45 (0.74, 2.84)	2.26/1	1.28 (0.64, 2.57)	1.88/1
Non-Hispanic White Female
No HIV	389	1,258,298	3.09 (2.8, 3.41)	Reference		Reference
HIV	32	25,928	12.34 (8.59, 17.21)	3.45 (1.97, 6.05)	6.36/3.35	3.52 (1.87, 6.63)	6.5/3.15
Non-Hispanic Black Female
No HIV	214	809,231	2.64 (2.31, 3.02)	Reference		Reference
HIV	143	82,621	17.31 (14.64, 20.33)	4.67 (3.65, 5.98)	8.81/6.76	4.42 (3.41, 5.73)	8.31/6.28
Non-Hispanic White Male
No HIV	264	802,294	3.29 (2.91, 3.71)	Reference		Reference
HIV	52	48,645	10.69 (8.07, 13.91)	2.76 (1.84, 4.14)	4.96/3.08	2.63 (1.74, 3.97)	4.7/2.87
Non-Hispanic Black Male
No HIV	138	396,909	3.48 (2.93, 4.09)	Reference		Reference
HIV	99	80,145	12.35 (10.09, 14.97)	2.62 (1.93, 3.54)	4.68/3.27	2.39 (1.76, 3.23)	4.21/2.92
Hispanic Male
No HIV	38	272,303	1.4 (1, 1.9)	Reference		Reference
HIV	12	19,168	6.26 (3.39, 10.64)	2.08 (0.95, 4.53)	3.58/1	1.98 (0.89, 4.39)	3.37/1
Age 50–64 Years
Overall
No HIV	7023	3,232,247	21.73 (21.22, 22.24)	Reference		Reference
HIV	720	197,705	36.42 (33.83, 39.15)	1.51 (1.32, 1.73)	2.39/1.97	1.46 (1.26, 1.67)	2.28/1.83
Female
No HIV	3289	1,861,041	17.67 (17.08, 18.28)	Reference		Reference
HIV	253	73,077	34.62 (30.55, 39.09)	1.62 (1.28, 2.07)	2.62/1.88	1.50 (1.13, 1.99)	2.37/1.51
Male
No HIV	3734	1,371,205	27.23 (26.37, 28.12)	Reference		Reference
HIV	467	124,628	37.47 (34.19, 40.99)	1.32 (1.12, 1.56)	1.97/1.49	1.31 (1.11, 1.55)	1.95/1.46
Non-Hispanic White
No HIV	3854	1,362,101	28.29 (27.41, 29.2)	Reference		Reference
HIV	171	42,215	40.51 (34.77, 46.93)	1.24 (0.99, 1.56)	1.79/1	1.21 (0.96, 1.53)	1.71/1
Non-Hispanic Black
No HIV	1910	751,715	25.41 (24.29, 26.57)	Reference		Reference
HIV	414	99,332	41.68 (37.81, 45.84)	1.64 (1.36, 1.97)	2.66/2.06	1.59 (1.35, 1.88)	2.56/2.04
Hispanic
No HIV	318	429,992	7.4 (6.62, 8.24)	Reference		Reference
HIV	40	16,825	23.77 (17.22, 32.06)	1.88 (1.12, 3.15)	3.17/1.49	1.75 (1.06, 2.88)	2.9/1.31
Non-Hispanic White Female
No HIV	1824	763,039	23.9 (22.83, 25.02)	Reference		Reference
HIV	53	12,046	44 (33.29, 57.11)	1.29 (0.80, 2.07)	1.9/1	1.23 (0.64, 2.35)	1.76/1
Non-Hispanic Black Female
No HIV	844	428,819	19.68 (18.39, 21.04)	Reference		Reference
HIV	157	39,538	39.71 (33.85, 46.3)	1.98 (1.56, 2.52)	3.37/2.49	1.89 (1.49, 2.40)	3.19/2.34
Non-Hispanic White Male
No HIV	2030	599,062	33.89 (32.44, 35.38)	Reference		Reference
HIV	118	30,169	39.11 (32.52, 46.67)	1.16 (0.88, 1.53)	1.59/1	1.15 (0.87, 1.52)	1.57/1
Non-Hispanic Black Male
No HIV	1066	322,896	33.01 (31.08, 35.04)	Reference		Reference
HIV	257	59,794	42.98 (37.96, 48.48)	1.31 (1.04, 1.65)	1.95/1.24	1.32 (1.07, 1.63)	1.97/1.34
Hispanic Male
No HIV	159	166,186	9.57 (8.16, 11.14)	Reference		Reference
HIV	28	10,991	25.48 (17.26, 36.33)	1.70 (0.96, 3.04)	2.79/1	1.66 (0.92, 3.01)	2.71/1

^aEvents per 10,000 person-years. ^b IPW; weight (1) ^c IPW; weight (2) Both weights were used to control for confounding and informative right censoring. The censoring model for both included time-updated age, calendar year, sex, race/ethnicity, US state, and comorbidity count. For the HIV model, both weights included baseline age, calendar year, sex, race/ethnicity, and US state; weight 2 additionally included baseline comorbidity count, while weight 1 did not.Age-stratified results for Hispanic female beneficiaries were not shown because we have limited case counts within stratums

The background mortality rate was higher among PWH as compared to PWoH overall and when stratified by sex and race/ethnicity (Table 3). Post-lung cancer diagnosis mortality increased substantially among both PWH and PWoH. Excess mortality was higher for PWH as compared to PWoH. Findings were similar when stratified by sex and race/ethnicity, though only significant (as reflected by CIs) among male and non-Hispanic Black beneficiaries. After adjusting for baseline comorbidity status, the pattern still held, except the excess mortality ratio was farther away from null (Table S2, Additional file 3).

Table 3.

Excess mortality overall and by HIV status, age, sex, and race/ethnicity among Medicaid beneficiaries from 14 states, 2001–2015

	Background Mortalitya			Post-Lung Cancer Diagnosis Mortalityb			Post-Diagnosis versus Background
	Observed Deaths	Observed Person-Years	Weighted Mortalityc (95% CI)	Observed Deaths	Observed Person-Years	Weighted Mortalityc (95% CI)	Weighted Excess Mortalityc (95% CI)	Ratio of Weighted Excess Mortalityc
Overall
Across all ages
No HIV	36,472	8,110,847	0.59 (0.58, 0.6)	2244	11,031	22.41 (20.68, 24.06)	21.82 (20.08, 23.48)	Reference
HIV	9325	529,150	1.76 (1.73, 1.8)	409	1423	28.74 (25.42, 32.21)	26.98 (23.65, 30.45)	1.24 (1.06, 1.43)
Age 30–49 Years
No HIV	11,560	4,878,616	0.35 (0.34, 0.36)	240	1789	14.83 (11.87, 18.63)	14.48 (11.52, 18.28)	Reference
HIV	5672	331,446	1.71 (1.66, 1.76)	132	453	29.12 (23.49, 35.99)	27.4 (21.77, 34.29)	1.89 (1.37, 2.63)
Age 50–64 Years
No HIV	24,912	3,231,602	0.86 (0.84, 0.88)	2004	9242	22.24 (20.06, 24.7)	21.38 (19.2, 23.84)	Reference
HIV	3653	197,551	1.85 (1.79, 1.91)	277	970	28.56 (24.93, 32.69)	26.71 (23.05, 30.85)	1.25 (1.05, 1.46)
Female
Across all ages
No HIV	16,973	4,923,815	0.42 (0.42, 0.44)	1046	5853	18.72 (16.73, 21.21)	18.3 (16.3, 20.79)	Reference
HIV	3677	220,192	1.67 (1.61, 1.72)	168	705	23.83 (19.33, 28.66)	22.16 (17.69, 27)	1.21 (0.93, 1.52)
Age 30–49 Years
No HIV	5449	3,062,783	0.24 (0.24, 0.25)	131	1052	13.96 (10.58, 18.33)	13.72 (10.33, 18.08)	Reference
HIV	2475	147,116	1.68 (1.62, 1.75)	78	247	31.59 (23.89, 41.45)	29.91 (22.21, 39.8)	2.18 (1.46, 3.21)
Age 50–64 Years
No HIV	11,524	1,860,671	0.65 (0.63, 0.67)	915	4801	17.95 (15.55, 20.85)	17.3 (14.9, 20.18)	Reference
HIV	1202	73,011	1.65 (1.55, 1.73)	90	458	19.64 (15.14, 25.33)	18 (13.51, 23.67)	1.04 (0.77, 1.43)
Male
Across all ages
No HIV	19,499	3,187,032	0.69 (0.68, 0.7)	1198	5178	23.71 (21.53, 26.46)	23.02 (20.84, 25.77)	Reference
HIV	5648	308,957	1.83 (1.78, 1.88)	241	718	33.56 (28.42, 39.97)	31.73 (26.58, 38.13)	1.38 (1.12, 1.72)
Age 30–49 Years
No HIV	6111	1,815,832	0.42 (0.4, 0.43)	109	737	15.15 (11.58, 19.22)	14.73 (11.16, 18.81)	Reference
HIV	3197	184,330	1.73 (1.67, 1.79)	54	206	26.15 (18.01, 36.4)	24.42 (16.27, 34.67)	1.66 (1.02, 2.75)
Age 50–64 Years
No HIV	13,388	1,370,931	0.99 (0.96, 1.01)	1089	4441	24.16 (21.54, 26.98)	23.18 (20.57, 25.99)	Reference
HIV	2451	124,539	1.97 (1.89, 2.05)	187	512	36.55 (30.53, 43.31)	34.58 (28.56, 41.35)	1.49 (1.19, 1.89)
Non-Hispanic White
Across all ages
No HIV	19,627	3,422,687	0.67 (0.66, 0.69)	1316	5738	23.1 (21.29, 25.19)	22.43 (20.62, 24.53)	Reference
HIV	1873	116,787	1.6 (1.53, 1.68)	89	307	28.99 (21.8, 37.92)	27.39 (20.19, 36.31)	1.22 (0.89, 1.65)
Age 30–49 Years
No HIV	6272	2,060,591	0.42 (0.41, 0.44)	143	888	16.49 (12.29, 21.59)	16.07 (11.88, 21.17)	Reference
HIV	1150	74,573	1.54 (1.45, 1.62)	24	95	25.24 (14.83, 38.69)	23.7 (13.3, 37.13)	1.47 (0.79, 2.47)
Age 50–64 Years
No HIV	13,355	1,361,769	0.96 (0.93, 0.99)	1173	4849	22.83 (20.55, 25.22)	21.87 (19.6, 24.27)	Reference
HIV	723	42,190	1.71 (1.6, 1.83)	65	212	30.67 (23.56, 40.13)	28.96 (21.83, 38.41)	1.32 (0.98, 1.81)
Non-Hispanic Black
Across all ages
No HIV	10,952	1,957,851	0.67 (0.65, 0.68)	668	3054	24.02 (21.32, 27.16)	23.36 (20.65, 26.5)	Reference
HIV	5511	262,097	2.1 (2.05, 2.16)	257	828	31.04 (26.15, 36.6)	28.94 (24.04, 34.5)	1.24 (1.01, 1.52)
Age 30–49 Years
No HIV	3571	1,206,140	0.39 (0.37, 0.4)	79	523	17.14 (12.45, 23.04)	16.75 (12.06, 22.65)	Reference
HIV	3414	162,766	2.1 (2.02, 2.17)	90	263	34.25 (25.54, 43.44)	32.15 (23.45, 41.37)	1.92 (1.27, 2.93)
Age 50–64 Years
No HIV	7381	751,545	0.96 (0.93, 1)	589	2531	23.54 (20.48, 27.44)	22.58 (19.5, 26.48)	Reference
HIV	2097	99,237	2.11 (2.03, 2.2)	167	565	29.55 (24.67, 35.32)	27.44 (22.53, 33.21)	1.22 (0.92, 1.54)

^aAll-cause mortality among beneficiaries who had not yet developed lung cancer

^bAll-cause mortality rate after lung cancer diagnosis

^cWeighting for baseline HIV status and right censoring (IPW; weight 1). The HIV model included baseline age, year, sex, race/ethnicity, and US state. The censoring model included time-updated age, calendar year, sex, race/ethnicity, US state, and time-updated comorbidity count. For results using weight 2, see Table S2 in the online supplemental materials. We were limited by event counts to report excess mortality among Hispanic beneficiaries

Lung cancer incidence and mortality: beneficiaries 30–49 years

Among those aged 30–49, crude incidence rates of lung cancer were higher among PWH than PWoH overall and when stratified by sex and race/ethnicity (Table 2). When stratified by calendar period (Fig. 1), incidence decreased over time but remained higher among PWH overall and among sub-groups. Male PWoH had higher lung cancer incidence than female PWoH, however, this difference was modest. In contrast, female PWH had slightly higher incidence than male PWH (Women: 13.53 per 10,000 person-years (11.74, 15.51); Men: 10.36 per 10,000 person-years (8.97, 11.91); Table 2).

Fig. 1.

Calendar time trends of lung cancer incidence by HIV status, age, sex, and race/ethnicity

The hazard of lung cancer incidence was 2–4 times higher among PWH aged 30–49 as compared to PWoH aged 30–49 overall and when further stratified by sex and all race/ethnicities, except for Hispanic beneficiaries (Table 2). Similar patterns were observed for the cumulative incidence of lung cancer (Figures S1-S6, Additional Files 4–9) and RD at ages 35 and 45 (Table 4). E-values suggest a strong unmeasured confounder would be needed to account for these findings (Table 4).

Table 4.

Cumulative incidence (risk%) of lung cancer by HIV status among Medicaid beneficiaries from 14 states, 2001–2015

Age	Risk, with HIV (95% CI)	Risk, without HIV (95% CI)	RD (95% CI)	RR (95% CI)	E-value for RR (point estimate/CI)
Crude
35	0.18 (0.09, 0.26)	0.03 (0.02, 0.03)	0.15 (0.07, 0.23)	7.09 (3.72, 11.21)	13.66/6.91
45	0.96 (0.83, 1.11)	0.24 (0.22, 0.26)	0.72 (0.59, 0.87)	4.03 (3.41, 4.75)	7.53/6.27
55	2.76 (2.55, 2.96)	1.3 (1.25, 1.34)	1.46 (1.26, 1.66)	2.13 (1.97, 2.31)	3.68/3.34
65	5.07 (4.77, 5.37)	3.42 (3.35, 3.49)	1.64 (1.35, 1.95)	1.48 (1.39, 1.57)	2.32/2.14
Weighteda
35	0.14 (0.05, 0.27)	0.02 (0.01, 0.02)	0.12 (0.04, 0.25)	7.01 (2.61, 13.77)	13.5/4.66
45	0.71 (0.56, 0.88)	0.21 (0.19, 0.23)	0.5 (0.35, 0.66)	3.33 (2.66, 4.2)	6.12/4.76
55	2.34 (2.07, 2.64)	1.26 (1.21, 1.31)	1.08 (0.79, 1.39)	1.86 (1.62, 2.11)	3.12/2.63
65	4.31 (3.89, 4.79)	3.34 (3.25, 3.43)	0.97 (0.53, 1.45)	1.29 (1.16, 1.43)	1.91/1.59
Weightedb
35	0.13 (0.05, 0.28)	0.02 (0.02, 0.03)	0.11 (0.03, 0.25)	6.7 (2.47, 13.51)	12.88/4.37
45	0.68 (0.54, 0.87)	0.21 (0.19, 0.23)	0.46 (0.32, 0.65)	3.16 (2.48, 4.11)	5.77/4.39
55	2.24 (1.94, 2.57)	1.27 (1.21, 1.31)	0.97 (0.67, 1.3)	1.77 (1.53, 2.04)	2.93/2.42
65	4.23 (3.8, 4.74)	3.35 (3.26, 3.44)	0.88 (0.44, 1.39)	1.26 (1.13, 1.42)	1.84/1.52

Abbreviations: RD risk difference, RR risk ratio

^aIPW; weight (1) ^b IPW; weight (2) Both weights were used to control for confounding and informative right censoring. The censoring model for both included time-updated age, calendar year, sex, race/ethnicity, US state, and comorbidity count. For the HIV model, both weights included baseline age, calendar year, sex, race/ethnicity, and US state; weight 2 additionally included baseline comorbidity count, while weight 1 did not

Among those aged 30–49, the background mortality rate was higher among PWH as compared to PWoH overall and when stratified by sex and race/ethnicity (Table 3). Post-lung cancer diagnosis mortality increased substantially among both PWH and PWoH. Excess mortality was higher for PWH as compared to PWoH. Relative excess mortality was higher among female beneficiaries (excess mortality ratio: 2.18; 95% CI:1.46, 3.21) as compared to male beneficiaries (excess mortality ratio:1.66; 95% CI:1.02, 2.75), although the CIs overlapped. Findings were similar when stratified by race/ethnicity, though only significant (as reflected by the CIs) among non-Hispanic Black beneficiaries.

Lung cancer incidence and mortality: beneficiaries 50–65 years

Lung cancer incidence was substantially higher among beneficiaries 50–65 years old as compared to beneficiaries 30–49 for both PWH and PWoH overall and when stratified by sex and race/ethnicity. Among those aged 50–65, the crude incidence rates of lung cancer were higher among PWH than PWoH overall and when stratified by sex and race/ethnicity (Table 2). When stratified by calendar period (Fig. 1), incidence decreased over time but remained higher among PWH overall and within sub-groups. Male PWoH had substantially higher lung cancer incidence than female PWoH. In contrast, the difference by sex was much less pronounced among PWH (Table 2). Notably, female PWH aged 50–65 were the only subgroup that did not experience a decline in lung cancer incidence in the most recent calendar period (Fig. 1).

The hazard of lung cancer was approximately 1.2–1.9 times higher among PWH aged 50–64 as compared to PWoH aged 50–64 overall and when stratified by sex and all race/ethnicities (Table 2); however, the HRs were lower among those 50–65 as compared to their 30–49-year-old counterparts overall and among all sub-groups except for Hispanic beneficiaries. Similar patterns were observed for the cumulative incidence of lung cancer (Figures S1-S6, Additional files 4–9) and RD at ages 55 and 65 (Table 4). E-values suggest a modest unmeasured confounder would be needed to account for these findings (Table 4).

Among those aged 50–65, the background mortality was higher among PWH as compared to PWoH overall and when stratified by sex and race/ethnicity (Table 3). Post-lung cancer diagnosis mortality increased substantially among both PWH and PWoH. Excess mortality was higher for PWH as compared to PWoH overall and among male beneficiaries. Relative excess mortality did not differ by HIV status among female beneficiaries (excess mortality ratio: 1.04; 95% CI: 0.77, 1.43). Relative excess mortality was higher among PWH as compared to PWoH when stratified by race/ethnicity; however, the CIs included 1.

Discussion

In nearly 4 million Medicaid beneficiaries aged 30–64, lung cancer incidence increased substantially with age regardless of HIV status, with PWH showing consistently higher incidence and hazards than PWoH. While crude lung cancer rates declined over time for both groups, males showed greater improvement, with minimal decrease among older female PWH. The elevated lung cancer hazard among PWH was strongest in younger beneficiaries. Both PWH and PWoH experienced increased mortality following lung cancer diagnosis, but PWH had greater excess mortality. Notably, younger female beneficiaries showed the strongest associations between HIV and both lung cancer hazards and excess mortality.

Our findings of an elevated risk of lung cancer among PWH compared to PWoH align with previous research findings [1, 3, 22, 23, 45–51]. We observed higher crude incidence rates compared to earlier studies [49–51], including the HIV/AIDS Cancer Match Study, likely due to both population and methodological differences. Our Medicaid cohort is enriched for lung cancer risk factors, including higher smoking prevalence, poverty, disability, and comorbidities [29, 52, 53], whereas the HIV/Cancer Match population likely more closely reflects the general population. Additionally, our methods differ: we drew both case counts and person-time from Medicaid data, censoring at loss of Medicaid coverage or death to capture only person-time of active healthcare observation. Despite applying washout periods to exclude those enrolling due to cancer diagnosis, our numerator may still include some recently diagnosed prevalent cases. While the HIV/Cancer Match Study may identify incident cases more accurately through cancer registry linkage, they use broader census-based person-time estimates. These combined differences in study populations and methods likely explain our higher observed crude rates.

It is well-established that smoking is one of the strongest risk factors for lung cancer development [54–56]. The association between HIV and lung cancer may be partly due to the high smoking prevalence among PWH [27, 57, 58]. It is estimated that 31% of all smoking-related cancers are attributable to smoking greater than 5 pack-years among PWH and that 94% of lung cancer cases could potentially be prevented by smoking cessation [57, 59]. Beyond the impact of smoking, population-based studies indicate that even after adjusting for smoking and age, PWH still showed 2–3 times increased risk of lung cancer, suggesting HIV infection may be independently associated with lung cancer risk [4, 12, 47, 59, 60]. Several hypotheses for how HIV may contribute to lung cancer incidence include the carcinogenic potential of HIV as a virus on its own, HIV-associated immunodeficiency, altered gene expression enhancing malignancy development, and HIV-related chronic inflammation [12]. Elevated risk may also partly reflect increased disease detection resulting from PWH’s potentially more frequent healthcare interactions [60]. However, current evidence suggests that lung cancer screening uptake is low overall, and limited evidence among PWH suggests similarly low screening uptake and follow-up [61–65].

We also observed excess mortality associated with lung cancer for PWH as compared to PWoH, which is consistent with two recent studies that evaluated excess mortality among PWH in the US [15, 66]. Our findings support previous evidence regarding decreased survival and worse overall prognosis among PWH with lung cancer in the modern antiretroviral therapy (ART) era [12, 14, 51]. Some evidence suggested that PWH tend to experience shorter post-diagnostic median survival compared to those without HIV, which may result from a higher proportion of advanced-stage diagnoses [7, 67]. However, other studies report no significant differences in cancer stage distributions according to HIV status [11, 12, 14]. Treatment inequities may also contribute to survival differences; large observational studies reported that PWH in the modern ART era were less likely to receive lung cancer treatment [13, 68]. Future studies should investigate the association between HIV and lung cancer treatment to provide additional insights into the observed survival disparities.

Importantly, we observed a notable sex- and age-based difference in lung cancer incidence and mortality. Few studies have investigated sex- and age-specific incidence of lung cancer among PWH. We observed that while the crude incidence rates of lung cancer declined over time in both PWH and PWoH – consistent with the consensus regarding decreasing incidence in the general population, and with recent evidence suggesting downward trends among PWH [22, 23] – the smallest absolute improvement observed was among older female PWH, whose rates remained relatively stable across the study periods. Notably, during 2011–2015, the crude rates (reflecting real-world burden) among older female PWH surpassed those observed in older male PWH. Further, the relative differences in lung cancer hazard and cumulative incidence by HIV status, after controlling for confounders, differential follow-up, and competing risks of death, were also more pronounced among female beneficiaries, particularly below the age of 50. We also observed greater excess mortality associated with lung cancer only in younger female PWH. These findings collectively underscore the public health importance of the disproportionate lung cancer burden (crude) and risk (after adjustment) among female PWH when developing mitigation strategies.

The reason for the observed sex difference in lung cancer incidence and all-cause mortality post-lung cancer diagnosis remains unclear. One possible explanation is differences in smoking behavior [21, 69]. In the US general population, men historically smoked more than women [70]. Among PWH in the US, previous studies show similar or higher smoking rates among men with HIV compared to women with HIV [27, 71, 72]. Furthermore, Medicaid enrollees generally smoked more than the general population (for both sexes) [28, 29], and therefore, the relative difference in smoking by sex may be reduced in Medicaid. Taken together, these findings indicate that smoking proportion alone is unlikely to explain the greater lung cancer disparities observed among female PWH in our Medicaid cohort. Increased exposure to second-hand smoke in women, especially at earlier stages in life, may play a role [16, 73–75]. Some evidence also suggests that sex differences in susceptibility to DNA damage, immune responses, and sex hormones could contribute to differences in incidence [16, 73, 76–80]. Moreover, lung cancer in women has broader and unique etiologies beyond smoking, warranting additional research [75]. Understanding mechanisms contributing to sex and age differences in lung cancer among PWH is essential for developing targeted intervention strategies.

The presence of HIV adds complexity to lung carcinogenesis and contributes to worse survival outcomes [11–14, 67]. Nevertheless, current lung cancer screening recommendations determine eligibility based primarily on age and smoking status for high-risk populations [81, 82]. As a result, some women at risk may be under-captured by existing criteria, since women can bear a greater lung cancer burden at younger ages in certain settings: for example, incidence for non-small cell lung cancer peaks earlier in women than in men [83], and among younger US adults, lung cancer incidence has surpassed that in men among non-Hispanic White and Hispanic women in recent birth cohorts [18]; women also tend to have lower smoking exposure and are overrepresented among lung cancer cases in never-smokers [16, 84]. The U.S. Preventive Services Task Force (USPSTF) and CMS have recently lowered the recommended screening age from 55 to 50 [81, 82], which, as suggested by simulation studies, may expand eligibility for some women [85]. Yet, sex-based gaps in screening may persist, as some women may still develop lung cancer with fewer pack-years and therefore remain ineligible despite the lower age threshold. These concerns may be especially relevant for PWH, for whom screening guidance remains limited [86, 87].

This study has several strengths. We evaluated a large cohort of PWH and PWoH enrolled in Medicaid across 14 US states, drawn from the same source, enabling comprehensive subgroup analyses. We accounted for competing risk of death and used weighted analyses to address differences in age, loss-to-follow-up, comorbidities, and other characteristics between groups. Comparing PWH and PWoH within Medicaid minimized differences in socioeconomic status, healthcare access, and smoking rates between HIV groups.

However, our findings have some limitations. First, we lacked smoking data; although comparing PWH and PWoH within Medicaid reduced smoking prevalence differences, residual confounding remains possible. The e-values suggest unmeasured confounding could partially explain our observed associations, given smoking’s strong relationship with lung cancer [87]. Second, we could not assess lung cancer stage, histologic subtypes, or HIV-related biomarkers (CD4 count, viral load), limiting mechanistic insights. Third, we defined HIV as a time-fixed exposure at baseline using diagnosis codes, without accounting for viral load or immune status. Furthermore, our data cover the years 2001–2015, which may reduce the applicability of the findings to more recent years. However, these years provide important data for studying lung cancer incidence and mortality in PWH, especially given the large number of PWH with Medicaid coverage and the limited evidence on sex- and age-specific incidence in the literature. Finally, our findings may not generalize beyond Medicaid-enrolled PWH to those with other insurance or socioeconomic circumstances.

Despite these limitations, our findings reveal important disparities in lung cancer risk and outcomes by HIV status, sex, and age. Younger female PWH had higher lung cancer hazards and greater excess mortality post-diagnosis compared to PWoH, while older female PWH showed minimal improvement in crude rates over time despite declining trends in other groups. These findings suggest HIV care providers should consider earlier lung cancer screening for female PWH who may not meet traditional criteria, integrate risk assessment into routine care, and ensure rapid HIV-oncology coordination when cancer is diagnosed.

Conclusions

Our findings confirm that PWH have elevated lung cancer incidence and excess mortality compared to PWoH, particularly among female PWH under age 50. Despite overall decreasing lung cancer rates, improvement was minimal among older female PWH, suggesting a need for targeted smoking cessation and lung cancer screening in this group. As PWH age, lung cancer’s contribution to morbidity and mortality is increasing. Future research should include prospective cohorts with detailed smoking and HIV biomarker data, trials or emulation studies testing modified screening criteria for PWH (particularly lower age and smoking thresholds for women), and implementation studies of integrated HIV-cancer care models.

Crude incidence rates (per 10,000 person-years) of lung cancer by HIV status across the three calendar periods under study are shown in panels A-C. Panel A shows incidence trends by HIV status and age; Panel B, by HIV status, age, and sex; Panel C, by HIV status, age, and race/ethnicity. Data for Hispanic beneficiaries are not shown due to limited case counts within age-calendar period strata.

Abbreviations

PWH

People with HIV

PWoH

People without HIV

EMR

Excess mortality ratio

US

United States

CMS

Centers for Medicare and Medicaid Services

ICD-9

International Classification of Diseases, Ninth Revision (ICD-9)

ICD-10

International Classification of Diseases, Tenth Revision (ICD-10)

CI

Confidence interval

HR

Hazard ratio

CIF

Cumulative incidence function

RD

Risk difference

RR

Risk ratio

SES

Socioeconomic status

ART

Antiretroviral therapy

USPSTF

The U.S. Preventive Services Task Force

Authors’ contributions

CEJ and BL developed the concept and design of this study and supervised the research team throughout the project. YZ performed statistical analysis, created data visualizations, and drafted the initial manuscript. XX processed and managed raw claims data for analysis. CEJ, BL, and JER critically revised the manuscript. KLC, MP, KY, and ELW reviewed and edited the manuscript. ELW also oversaw project administration. All authors have read and approved the manuscript as submitted.

Funding

This work was supported by National Institutes of Health grants R01CA250851 andP30AI094189. 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 data that support the findings of this study are available from the Centers for Medicare and Medicaid Services (CMS) and administered by ResDAC. Investigators may reuse these data if they independently meet CMS requirements and obtain both CMS reuse approval and permission from the study’s National Institutes of Health program officer.

Declarations

Ethics approval and consent to participate

This study was consistent with the Helsinki Declaration. The Johns Hopkins Bloomberg School of Public Health Institutional Review Board reviewed this secondary analysis of Medicaid claims data and determined that it was exempt and waived the requirement for informed consent (IRB00011617).

Consent for publication

Not applicable.

Competing interests

The authors report no conflicts of interest.