Providers' mediating role for medication adherence among cancer survivors.

BACKGROUND: We conducted a mediation analysis of the provider team's role in changes to chronic condition medication adherence among cancer survivors.
METHODS: We used a retrospective, longitudinal cohort design following Medicare beneficiaries from 18-months before through 24-months following cancer diagnosis. We included beneficiaries aged ≥66 years newly diagnosed with breast, colorectal, lung or prostate cancer and using medication for non-insulin anti-diabetics, statins, and/or anti-hypertensives and similar individuals without cancer from Surveillance, Epidemiology, and End Results-Medicare data, 2008-2014. Chronic condition medication adherence was defined as a proportion of days covered ≥ 80%. Provider team structure was measured using two factors capturing the number of providers seen and the historical amount of patient sharing among providers. Linear regressions relying on within-survivor variation were run separately for each cancer site, chronic condition, and follow-up period.
RESULTS: The number of providers and patient sharing among providers increased after cancer diagnosis relative to the non-cancer control group. Changes in provider team complexity explained only small changes in medication adherence. Provider team effects were statistically insignificant in 13 of 17 analytic samples with significant changes in adherence. Statistically significant provider team effects were small in magnitude (<0.5 percentage points).
CONCLUSIONS: Increased complexity in the provider team associated with cancer diagnosis did not lead to meaningful reductions in medication adherence. Interventions aimed at improving chronic condition medication adherence should be targeted based on the type of cancer and chronic condition and focus on other provider, systemic, or patient factors.

Introduction

More than 60% of Medicare beneficiaries diagnosed with cancer also have three or more chronic conditions [1]. Management of chronic conditions among cancer survivors is complex [2-5], especially medication management [6]. Increasing evidence suggests that adherence to medications for chronic conditions decreases in older adults with some cancers [7-13]. For example, our earlier study found that adherence to anti-diabetics and statins declined among older adults with colorectal and lung cancer by two to four percentage points after cancer diagnosis relative to matched non-cancer patients [13].

However, little is known about the mechanisms for these changes in chronic condition medication adherence among cancer survivors. A diagnosis of cancer can directly affect medication adherence for other chronic conditions in a variety of ways. Cancer diagnosis can shift the emphasis of medical care to the emerging cancer. For example, in the presence of additional cost and complexity created by cancer-related prescriptions (both treatment and symptom management), patients may decrease adherence to medications for other chronic conditions. Conversely, cancer diagnosis may reinforce the importance of chronic disease prevention, serving as a “wake-up call” and encouraging healthy behaviors such as adherence to medications.

Another pathway for changes in adherence is through changes to the survivor’s provider team. After diagnosis, patients see new oncology specialists and may or may not continue to see their original primary care provider and chronic disease specialists even after cancer treatment, which can disrupt communication and coordination among the provider team [4, 5, 14]. Additional providers and medical visits may lead to differential clinical priorities or management strategies across multiple providers, confuse patient-provider communication, or make care coordination more difficult, which may influence chronic condition medication adherence [14].

Understanding the role of the provider team can help inform policy and practice aimed at improving care for patients with multiple chronic conditions. Alternative payment models increasingly put providers at financial risk for holistic patient outcomes, increasing the incentive to coordinate care for chronic conditions. Several quality improvement efforts to improve care for cancer survivors focus on transitioning care from oncology specialists back to primary care providers [15] and improving coordination of care among the provider team [14, 16–18].

In this study, we investigated the mechanisms for cancer-related changes in chronic condition medication adherence among older cancer survivors. Specifically, we conducted a mediation analysis to investigate the role of changes in the provider team structure in adherence changes during and after primary cancer treatment. We hypothesized that the increase in the number of providers and specialists and more complicated patient-sharing relationships will make coordination more difficult and lower medication adherence for chronic conditions [14, 16, 19].

Methods

Data source and study populations

We used the linked Surveillance, Epidemiology, and End Results program (SEER) cancer registries and Medicare enrollment and claims data [20]. The SEER registries collect demographic, tumor, and vital status data for incident cancers, covering approximately 34% of the United States population. Medicare enrollment and claims data record longitudinal information about healthcare utilization for beneficiaries enrolled in the fee-for-service program.

We identified patients aged ≥66 with a first primary diagnosis of stage I-III breast, prostate, non-small cell lung, or colorectal cancer from July 1, 2008 –December 31, 2012. Access to the data was originally granted in 2017. We excluded stage IV and metastatic disease for all cancer sites due to the different incentives patients and providers would face for chronic condition medication adherence considering limited life expectancy. Individuals diagnosed at autopsy or death were excluded. Individuals had to have Medicare Parts A, B, and D coverage for the 18-months before through 24-months following the month of cancer diagnosis.

For each condition (i.e., hyperlipidemia, hypertension, and diabetes mellitus), we constructed separate cohorts with at least one International Classification of Diseases, 9th Edition, Clinical Modification diagnosis code for the condition of interest and at least one prescription drug claim for an oral medication to manage that condition from -18 months to -7 months before cancer diagnosis. This approach resulted in 12 distinct cohorts; survivors could be represented in multiple cohorts.

For each cancer-chronic condition cohort, we identified a non-cancer comparison cohort using a 5% random sample of Medicare beneficiaries identified within each SEER region. For each cancer survivor (in each cancer-chronic condition cohort), we identified all potential individuals without a diagnosis of cancer who met the same chronic condition criteria as the cancer-chronic condition cohorts. We matched cancer and non-cancer individuals with the same chronic condition on exact age (in years), sex, race (White, Black, Asian, Hispanic, Native American Indian, Other), and SEER region. Among all eligible individuals, one non-cancer comparison patient was selected at random with replacement and assigned an index date, based on their matched cancer survivor’s diagnosis date. As in the cancer cohorts, controls could be in multiple condition cohorts [13].

Medication adherence

The primary outcome was chronic condition medication adherence, measured using the proportion of days covered (PDC) [21]. The PDC is the number of days covered by a prescription drug divided by the total number of days in an observation window. PDC has been shown to be more reliable than self-report [22, 23] and correlated with drug levels [24]. We removed hospitalizations and skilled nursing facility stays from the denominator and carrying forward any days’ supply which overlapped with a hospital or skilled nursing facility stay [21]. Adherence was evaluated at the condition- (e.g., hypertension) level; switching within and across drug classes was allowed.

The PDC was measured in 6-month time windows ending 1) at cancer diagnosis, 2) one year post-diagnosis (after which primary treatment is likely complete [25, 26]), and 3) two years post diagnosis. For the analysis of antidiabetics, we excluded all survivors that initiated insulin at any point during follow-up. PDC calculations are unreliable for insulin, and removal of survivors who initiate insulin is consistent with CMS specifications [27]. The PDC was dichotomized at ≥80% (adherent) versus <80% (non-adherent), a common cut-point [28-37]. As a secondary outcome, we defined discontinuation as a dichotomous indicator equal to one if the survivor did not fill a drug for their chronic condition for 90 continuous days [38, 39].

Provider team structure

We defined four measures of the provider team structure. We counted 1) the total number of providers and 2) number of specialists seen by each person in the period. All providers except the following were considered specialists: internal medicine doctors without subspecialty training (National Provider Identifier = 207R00000X), family practitioners (207Q00000X), general practitioners (208D00000X), obstetrics and gynecologists (207V00000X, 207VG0400X), and geriatricians (207RG0300X, 207QG0300X). For providers with more than one specialty, they were assigned to the specialty listed in most of their claims. We excluded mid-level providers (e.g., nurse practitioners) due to the variety of regulations on prescriptive authority across states. These measures captured the number of providers and specialists required to coordinate; more providers/specialists make care coordination more difficult [16].

Third, we defined degree as the count of all providers that share patients with the patient’s main provider [19]. We designated the main provider using the plurality provider algorithm [40, 41]. This provider team measure captured the level of coordination required between the main provider and all other providers; more provider peers make care coordination more difficult [19].

Fourth, for each pair of providers on a patient’s provider team, we calculated the proportion of each providers’ patients who were shared with the other provider in the dyad in the previous (6-month) period [14]. We counted a shared patient if two providers billed for outpatient evaluation and management visits. We then calculated the shared patient volume for a provider team using the geometric mean of all pairwise proportions. Shared patient volume was undefined for patients with only one provider; we set shared patient volume to zero in these cases. Shared patient volume captured the recent history of opportunities to coordinate for a patient’s provider team; higher levels of shared patient volume represent potentially higher degrees of care coordination.

Statistical analysis

Our causal model is presented in S1 Appendix. The following analyses were conducted separately in 24 analytic samples defined by combinations of cancer site (n = 4), chronic condition (n = 3), and time comparison (n = 2; diagnosis vs one year post-diagnosis and diagnosis vs two years post-diagnosis). We condensed the four provider team measures to two using factor analysis. We transformed the four provider team measures to be deviations from individual means by subtracting each patient’s mean over time from each variable (i.e., ). Factor analysis using the transformed provider measures consistently identified two factors (with positive eigenvalues) across all analysis samples. The first, “number of providers,” had high factor loadings for the number of total providers and number of specialists. The second factor, “sharing among providers,” had high factor loadings for degree and shared patient volume. We generated predicted values for these two factors and standardized each factor to be mean zero and in units of standard deviations.

For each analytic sample, we estimated three linear regressions. All regression variables were expressed as deviations from individual means over time to eliminate time-invariant patient characteristics (e.g., tumor features at diagnosis). The first two regressions had the provider team factors as dependent variables and cancer status as the explanatory variable. The third regression had adherence as the dependent variable with cancer and both provider team factors as explanatory variables. We used seemingly unrelated regression with robust standard errors to stack the within-person regressions into one variance-covariance matrix.

We report the following effects of cancer and provider team on adherence. The natural direct effect (NDE) is the expected difference in adherence between those with and without cancer holding the care team constant in the non-cancer configuration. The natural indirect effect (NIE) is the expected difference in adherence among cancer survivors comparing care teams with and without cancer. Total effect (TE) is the sum of NDE and NIE. See S1 Appendix for derivations.

All statistical analyses were performed using Stata version 15.1 (College Station, TX). This study was approved by the University of North Carolina at Chapel Hill Institutional Review Board.

Results

The analytic cohort has been described previously [13] and includes 11,831 unique individuals diagnosed with breast cancer, 6,580 with colorectal cancer, 4,105 with lung cancer and 11,879 men diagnosed with prostate cancer, each matched to a non-cancer control. Each cancer cohort experienced notable changes to their provider team after cancer diagnosis compared to their matched non-cancer control group (Table 1). For each cancer cohort, the number of providers and number of specialists increased one-year post-diagnosis relative to the non-cancer control group, which experienced smaller increases over time as the cohort aged. Similarly, each cancer cohort experienced a large increase in their main provider’s degree (number of other providers with whom they share patients) in the first year relative to the non-cancer controls. Patient’s main provider’s shared patient volume, although low overall (i.e., only about 1–2% of patients were shared on average amongst provider team), also increased in the first year for cancer survivors relative to non-cancer controls. The increase in patient volume shared among cancer patients is due to higher rates of patient sharing among oncologists, who often become cancer patients’ main provider. For all cancer cohorts and provider team variables, the averages decreased from the first year to the second year post-diagnosis for the cancer cohorts but remained higher than the non-cancer controls in the same time period. The full set of factor loadings are available in S1 Table.

Table 1

Provider team characteristics by cancer and time period: Mean (standard deviation).

Cancer	Timed	Factor: Numbera				Factor: Sharing
		Providers		Specialists		Degreeb		Shared Patient Volumec
		Cancer	Control	Cancer	Control	Cancer	Control	Cancer	Control
Breast(N = 11,831)	Diagnosis	1.99	1.84	1.26	1.18	51.43	49.87	0.014	0.014
		(1.57)	(1.65)	(0.64)	(0.64)	(57.04)	(59.59)	(0.049)	(0.052)
	1 year	3.60	1.92	2.13	1.19	102.06	52.74	0.022	0.014
		(1.99)	(1.78)	(0.72)	(0.65)	(100.69)	(61.78)	(0.047)	(0.053)
	2 years	3.21	1.97	1.97	1.20	86.86	53.24	0.018	0.014
		(2.09)	(1.85)	(0.77)	(0.66)	(96.53)	(64.02)	(0.046)	(0.051)
Colorectal(N = 6,580)	Diagnosis	2.60	1.92	1.60	1.24	61.90	53.44	0.014	0.014
		(2.11)	(1.76)	(0.79)	(0.70)	(63.10)	(61.44)	(0.045)	(0.052)
	1 year	3.34	1.99	1.89	1.24	101.05	55.48	0.021	0.015
		(2.48)	(1.92)	(0.79)	(0.71)	(96.51)	(63.88)	(0.047)	(0.053)
	2 years	2.99	2.09	1.75	1.26	85.95	56.87	0.018	0.014
		(2.45)	(2.10)	(0.80)	(0.71)	(92.37)	(67.34)	(0.048)	(0.051)
Lung(N = 4,105)	Diagnosis	2.96	1.93	1.71	1.24	74.67	54.24	0.013	0.014
		(2.36)	(1.86)	(0.81)	(0.70)	(79.28)	(63.98)	(0.046)	(0.048)
	1 year	3.98	2.04	2.11	1.25	115.14	56.73	0.020	0.015
		(2.74)	(2.04)	(0.72)	(0.72)	(104.39)	(66.26)	(0.044)	(0.057)
	2 years	3.88	2.15	2.05	1.27	106.19	58.46	0.016	0.014
		(3.02)	(2.34)	(0.78)	(0.73)	(109.21)	(69.37)	(0.042)	(0.050)
Prostate(N = 11,879)	Diagnosis	2.52	1.91	1.87	1.26	75.35	54.41	0.014	0.013
		(1.56)	(1.72)	(0.74)	(0.74)	(73.88)	(65.13)	(0.034)	(0.048)
	1 year	2.90	1.96	1.95	1.27	87.19	57.55	0.018	0.013
		(1.82)	(1.87)	(0.71)	(0.75)	(84.00)	(68.52)	(0.037)	(0.047)
	2 years	2.80	2.06	1.87	1.28	79.33	58.31	0.016	0.014
		(1.93)	(2.06)	(0.76)	(0.76)	(82.98)	(71.37)	(0.035)	(0.052)

a Providers and Specialists represent the total number of providers and number of specialists, respectively, seen by each person in the period.

b Degree is the count of all providers that share patients with the patient’s main provider.

c For each pair of providers on the patient’s team, we calculated the proportion of each providers’ patients who were shared with the other provider in the dyad in the previous period. Shared patient volume is the geometric mean all pairwise proportions.

d Six months ending at the time period indicated (relative to diagnosis date).

For non-insulin anti-diabetics, the proportion of adherent survivors decreased by 4 to 7 percentage points among those diagnosed with colorectal and lung cancer (Fig 1). There were no significant changes in anti-diabetic adherence among breast cancer survivors. Prostate cancer survivors experienced an increase in anti-diabetic adherence of 2 percentage points (95% Confidence Interval [CI]: 0.00–0.04) one year after diagnosis. A similar pattern was observed for adherence to statins with no significant changes among breast cancer survivors, 4 to 6 percentage point reductions among colorectal and lung cancer survivors, and a small increase of 1 percentage point (95% CI: 0.00–0.003) among prostate cancer survivors at both time points (Fig 2). The proportion of survivors adherent to anti-hypertensives increased among breast cancer survivors at both time points by 2 percentage points (95% CI: 0.02–0.03), increased by 2 percentage points (95% CI: 0.01–0.03) among colorectal cancer survivors at both time points, and increased by 4 percentage points (95% CI: 0.03–0.04) among prostate cancer survivors at both time points (Fig 3). Discontinuation rates increased among all cancer cohorts at both time points for anti-diabetics and statins but less so for anti-hypertensives (Fig 1 in S1 Fig).

Fig 1

Total effect of cancer on proportion adherent (proportion of days covered > 80%) for non-insulin anti-diabetics by cancer site and phase of care.

Point estimates and 95% confidence intervals.

Fig 2

Total effect of cancer on proportion adherent (proportion of days covered > 80%) for statins by cancer site and phase of care.

Point estimates and 95% confidence intervals.

Fig 3

Total effect of cancer on proportion adherent (proportion of days covered > 80%) for anti-hypertensives by cancer site and phase of care.

Point estimates and 95% confidence intervals.

Next, for each statistically significant TE, we tested for statistically significant NIE (combined across the two mediating factors) as evidence of mediation through the provider team. Only 4 of the 17 cancer/chronic condition/time combinations with significant TE had statistically significant NIE. Even in those 4 analytic cohorts, the NIE were practically very small. For example, among lung cancer survivors, adherence to statins decreased by 5 percentage points (95% CI: -0.08 –-0.03); the NDE was a 4.5 percentage point decrease (95% CI: -0.07 –-. 02) and the NIE represented only 16% (95% CI: 2–30) of the TE (Fig 4). Similarly, the significant increases in adherence to anti-hypertensives among colorectal cancer survivors (at two years post-diagnosis) and prostate cancer survivors were mostly due to NDE and not mediation through the provider team (Fig 5). The proportion of the TE mediated through the NIE was 18% (95% CI: 4–31) among colorectal cancer survivors, 6% (95% CI: 3–9) among prostate cancer survivors at one year and 3% (95% CI: 1–4) among prostate cancer survivors at two years.

Fig 4

Effects of cancer on proportion adherent (proportion of days covered > 80%) for statins: Total Effect (TE), Natural Direct Effect (NDE) and Natural Indirect Effect (NIE) through number of providers and sharing amongst providers.

Point estimates, in percentage point changes, for cancer sites and phases of care with statistically significant total and net indirect effect.

Fig 5

Effects of cancer on proportion adherent (proportion of days covered > 80%) for anti-hypertensives: Total Effect (TE), Natural Direct Effect (NDE) and Natural Indirect Effect (NIE) through number of providers and sharing amongst providers.

Point estimates, in percentage point changes, for cancer sites and phases of care with statistically significant total and net indirect effect.

For discontinuation, 10 of the 19 analytic cohorts with significant TE also had a statistically significant NIE (Fig 2 in S1 Fig). In each case, the NIE worked to offset the increase in discontinuation from the NDE (i.e., survivors who saw more providers/specialists and whose providers shared more patients were less likely to discontinue). However, the magnitude of the mediating effects (i.e., NIE) were small relative to the direct effect of cancer on discontinuation (i.e., NDE).

Discussion

A growing body of literature demonstrates significant changes in chronic disease medication adherence associated with cancer diagnosis and treatment [7-13]. The appropriate points for intervention to improve adherence, however, remain undefined. We found that changes in the provider team structure accompanying cancer diagnosis explained only a small portion of changes in chronic condition medication adherence. Provider team structure effects were statistically insignificant in 13 of 17 analytic samples with significant changes in medication adherence. When the provider team structure effects were statistically significant, they were small in magnitude (<0.5 percentage points). The results for provider team structure effects were similar for discontinuation.

What do these findings mean for policy and clinical interventions? As payers move toward value-based care, providers are assuming more financial risk for holistic patient outcomes. In bundled payment arrangements like the Oncology Care Model, providers are responsible for all care delivered to cancer patients on chemotherapy. For example, inpatient admissions for uncontrolled hypertension or diabetes will hurt quality metrics and, therefore, the shared savings amounts for participating providers. Thus, these arrangements provide incentives for providers to better coordinate care to improve adherence to chronic disease medications among cancer patients.

This is the first study, to our knowledge, to disentangle the influence of care coordination from other factors that might influence chronic medication adherence in oncology. Quality improvement efforts have emphasized care coordination during transitions from oncology specialty care to primary care providers [15] and improved care coordination across a patient’s provider team [14, 16–18]. However, our results suggest that the role of increased complexity in the provider team may be outsized by other factors that can contribute to poor medication adherence for comorbid chronic conditions. This highlights a need to better understand the role of other factors that may reduce medication adherence in oncology populations.

While this paper provides evidence about complexity of the provider team, several other factors may influence adherence to chronic medications including other provider characteristics (e.g., communication, reimbursement arrangements), social determinants of health, the financial burden of cancer treatment, and physiological responses to cancer and its treatments. Social determinants of health are the conditions in the environments where people live (e.g., transportation, stable housing) and have been linked to medication adherence [42]. The financial toxicity of cancer treatment is a known constraint for many patients [43, 44], which may translate beyond cancer into the management of comorbid chronic conditions. Additionally, the chemotherapeutic toxicity of many cancer treatments (e.g., emetogenic chemotherapy) can reduce a patient’s adherence to comorbid medications [45]. Furthermore, physiological responses may result in imbalances in blood pressure and blood glucose control, which can influence medication adherence [8, 46]. It is important to understand the range of factors that may influence chronic condition medication adherence as the interventions to improve each of these factors vary significantly.

This study had several strengths. First, we included survivors with one or more of several chronic conditions diagnosed with one of the four most common cancer sites [13]. Second, our estimation approach separates the effects of cancer diagnosis on medication adherence from underlying aging trends using a matched cohort of non-cancer controls. Our estimates also adjust for other factors influencing adherence that do not change over time (e.g., “healthy users,” health literacy, tumor stage). Finally, we report short- (one year) and long-run (two years) effects of cancer on chronic condition medication adherence beyond initial treatment.

This study also had limitations. First, medication adherence was evaluated using dispensed prescriptions and we cannot assume that all filled medications were consumed. Second, this study used data from 2008–2014 and was restricted to adults age 66 and older with continuous Medicare fee-for-service and Part D coverage with non-metastatic cancer who also survived two years following their cancer diagnosis. As such, our findings may not be generalizable to more current patients, those with Medicare Advantage or without prescription drug or other healthcare insurance, the population 65 years and under, patients with metastatic disease, or patients with a short life expectancy. Furthermore, we were not able to assess the effect of care coordination on mortality. Third, we cannot determine whether the observed changes in chronic condition medication adherence were clinically appropriate. Fourth, several mechanisms exist for the NDE of cancer (e.g., competing long-term cancer therapies like endocrine therapy) and more research is needed to unpack these mechanisms. Finally, causal interpretation of our results depends on the causal model and accompanying identification assumptions being correct.

This study found that changes in medication adherence due to cancer diagnosis differed across cancer sites and chronic conditions. The largest decreases in chronic condition medication adherence occurred for anti-diabetics and statins among colorectal and lung cancer survivors, while adherence to anti-hypertensives increased among breast, colorectal and prostate cancer survivors. The decreases in adherence among colorectal and lung cancer patients are consistent with a hypothesis that patients diagnosed with more deadly cancer have decreased incentive to manage chronic conditions, perhaps because preventive medication has a long lag-time for benefit [47] or because more complex treatments (e.g., surgical resection and adjuvant chemotherapy) change the need or benefits of continued use of chronic disease medications [13]. Conversely, patients diagnosed with cancers with high survival probabilities (e.g., breast and prostate) are motivated to improve their (secondary) prevention efforts, including medication adherence. The positive effects of cancer diagnosis on antihypertensive medication adherence may be explained by the fact that blood pressure monitoring is routine during cancer care visits, which provides ample opportunity for providers to promote the importance of antihypertensive adherence. Alternatively, monitoring of lipid and blood sugar levels is less routine, which might partially explain the decreases in adherence to these medications among cancer patients. Of course, other explanations are also plausible.

For all cancer sites and chronic conditions, cancer diagnosis led to increased number of providers, specialists and patient sharing among the provider team. However, the increased complexity in the provider team structure associated with cancer diagnosis did not lead to meaningful changes in medication adherence for chronic conditions. These results suggest that policies and interventions aimed at improving chronic condition medication adherence need to be targeted based on the type of cancer and chronic condition and can focus on systemic and patient factors that are present across provider teams for greater effect.

Details of causal model.

(DOCX)

Click here for additional data file.

Results for discontinuation outcome.

(DOCX)

Click here for additional data file.

Factor weights by cancer site and chronic condition.

(DOCX)

Click here for additional data file.

14 Jun 2021

PONE-D-21-09579

Providers’ Mediating Role for Medication Adherence among Cancer Survivors

PLOS ONE

Dear Dr. Trogdon,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please pay special attention to the concerns raised by reviewer 2 on the discussion

Please submit your revised manuscript by Jul 12 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Ilana Graetz

Academic Editor

PLOS ONE

Journal Requirements:

1) Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

2) Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

3) In the ethics statement in the manuscript and in the online submission form, please provide additional information about the patient records used in your retrospective study, including the date range (month and year) during which patients' medical records were accessed.

4)  We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. Please see http://www.bmj.com/content/340/bmj.c181.long for guidelines on how to de-identify and prepare clinical data for publication. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

5)  We note that the grant information you provided in the ‘Funding Information’ and ‘Financial Disclosure’ sections do not match.

When you resubmit, please ensure that you provide the correct grant numbers for the awards you received for your study in the ‘Funding Information’ section.

6) Thank you for stating the following in the Competing Interests section:

[Dr. Lund’s spouse is a full-time, paid employee of GlaxoSmithKline who also holds

stock in the amount of approximately $42,000. Dr. Lund also receives unrelated grant

funding paid to her institution from AbbVie. Dr. Wheeler receives unrelated grant

funding paid to her institution from Pfizer. All other co-authors have no potential

competing interests to report.].

Please confirm that this does not alter your adherence to all PLOS ONE policies on sharing data and materials, by including the following statement: "This does not alter our adherence to  PLOS ONE policies on sharing data and materials.” (as detailed online in our guide for authors http://journals.plos.org/plosone/s/competing-interests).  If there are restrictions on sharing of data and/or materials, please state these. Please note that we cannot proceed with consideration of your article until this information has been declared.

Please include your updated Competing Interests statement in your cover letter; we will change the online submission form on your behalf.

Please know it is PLOS ONE policy for corresponding authors to declare, on behalf of all authors, all potential competing interests for the purposes of transparency. PLOS defines a competing interest as anything that interferes with, or could reasonably be perceived as interfering with, the full and objective presentation, peer review, editorial decision-making, or publication of research or non-research articles submitted to one of the journals. Competing interests can be financial or non-financial, professional, or personal. Competing interests can arise in relationship to an organization or another person. Please follow this link to our website for more details on competing interests: http://journals.plos.org/plosone/s/competing-interests

7) Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This is a mediation analysis of the provider team’s role in changes to chronic condition medication adherence among cancer survivors. The study topic is of interest from both clinical and policy perspective. Further, the study has a reasonable methodology, although has limitations in terms of using SEER-Medicare data and relying on pharmacy refills as a measure of adherence. There are few minor comments regarding this study:

1- It is understandable that the authors have focused on 4 most common cancers, but it is not clear why in each set of cancers they have limited their population. For example for breast cancer, why stage IV was not included? or in lung cancer, why small cell lung was not included. If there is concern about heterogeneity in management, similar heterogeneity can be expected with colorectal cancer stages as well.

2- One of the main limitation of the study is that the data is from a decade ago, and might not be generalizable to current practices.

3- In assigning provider team structure, there has been no mention of mid levels who commonly manage patients as primary care. Why mid levels were excluded from the study?

4- It is not clear why patient sharing with other providers, or count of all providers sharing patients (degree) is increasing among cancer patients vs. controls (in Table 1). Why a single patient cancer diagnosis impacting the patient volume shared?

5- Authors may want to expand on why they think there is differences in adherence based on the chronic condition (why adherence to anti hypertensives higher) or type of cancer. Also there are controversial publications in this area; for example some other papers have shown breast cancer patients will have decreased adherence to chronic medications (as opposed to this study).

6- In supplemental appendix, there is an error displaying Fig 2A-C, and the images could not be visualized.

Reviewer #2: The reviewer has high enthusiasm for this well designed and well written analysis assessing the role of the complexity of the provider team on adherence to chronic disease medications for cancer survivors. This study includes many strengths, including looking at several different cancers and using robust methods with several novel provider complexity metrics. It also offers some good news: that our healthcare system can accommodate increasing provider complexity without risking patient adherence outcomes. These comments and questions are offered in the spirit of improving the clarity of the methods and carefulness of the language in the discussion.

Methods

� At first I was concerned about the exclusion about insurance, as those experience insurance losses may be most vulnerable to adherence challenges. At the same time, I believe this was a strong analytic choice, as it’s important to show that insurance changes were not the cause of non-adherence, and allows the team to further isolate the role of provider complexity.

� Was there censorship due to death? That is to say, that those with poorly coordinated care might have shorter survival. I think not, given that the limitations suggest that only those surviving two years are excluded. I think this is a point worth extracting out a bit: that it’s possible that provider complexity may have been important to survival overall, but, by design, that was not assessed in this analysis.

� I’m not familiar with the term “demeaning”. Is this similar to group mean centering or grand mean centering?

� It seems that this analysis has time nested in patients nested in providers. Were hierarchical models used? If no hierarchical models were used, how would you account for shared variance due to clustering of patients within providers?

� Some of the cancers include long-term therapies that might also compete with comorbidity management. For example, endocrine therapy for breast cancer may be last up to 5 years, and include both oral and intravenous components. Were those medications also accounted for in some way?

Results

� The mediation model adds value, but I also wonder if there are moderating effects. Given the results in the figures showing differences by tumor site, have you explored whether the interaction of provider complexity with tumor site might lead to differences in adherence?

Discussion

� Some of the language in the discussion extends beyond what the results of this study suggest, especially with the language about the need to focus on patient-level factors. Currently, it reads as if patient self-efficacy is the culprit for adherence challenges, as if the problem is that patients just aren’t motivated enough to be adherent. It comes across as victim-blaming and as if increasing a sense of personal responsibility will fix adherence issues. Current interventions for medication adherence for cancer survivors that focus on patient-level factors suggest otherwise1-5. If we can reasonably assume that patients do care about their health to some degree, there are likely competing priorities that may be wholly outside of health care that patients are busy addressing and influences their adherence. While this paper providers strong evidence about complexity of the provider team, there may be other provider-level factors that are still important that are not explored here. It is premature to suggest we should switch to focusing on patient factors. I recommend this language on p14 be revised substantially.

� One of the values of this paper is that it gives us a start on the evidence around the influence of providers: but also leaves more room to explore. For example, might provider complexity influence disparities in adherence? Might there be rural-urban differences to consider or issues of geography? Rather than saying we should immediately shift to patient-level factors, it would be worthwhile to point to other provider-level issues to explore.

References

1. Neven P, Markopoulos C, Tanner M, et al. The impact of educational materials on compliance and persistence rates with adjuvant aromatase inhibitor treatment: First-year results from the Compliance of ARomatase Inhibitors AssessmenT In Daily practice through Educational approach (CARIATIDE) study. The Breast. 2014;23(4):393-399.

2. Ziller V, Kyvernitakis I, Knöll D, Storch A, Hars O, Hadji P. Influence of a patient information program on adherence and persistence with an aromatase inhibitor in breast cancer treatment-the COMPAS study. BMC cancer. 2013;13(1):407.

3. Hadji P, Blettner M, Harbeck N, et al. The Patient's Anastrozole Compliance to Therapy (PACT) Program: a randomized, in-practice study on the impact of a standardized information program on persistence and compliance to adjuvant endocrine therapy in postmenopausal women with early breast cancer. Annals of oncology. 2013;24(6):1505-1512.

4. Lambert LK, Balneaves LG, Howard AF, Gotay CC. Patient-reported factors associated with adherence to adjuvant endocrine therapy after breast cancer: an integrative review. Breast cancer research and treatment. 2018;167(3):615-633.

5. Greer JA, Amoyal N, Nisotel L, et al. A systematic review of adherence to oral antineoplastic therapies. The oncologist. 2016;21(3):354-376.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

4 Oct 2021

Thank you for the comments. We provide a detailed response to each comment in the attached Response to Reviewers (PLOS_One_response_reviewers_final_v2.docx).

Submitted filename: PLOS_One_response_reviewers_final_v2.docx

Click here for additional data file.

9 Nov 2021

Providers’ mediating role for medication adherence among cancer survivors

PONE-D-21-09579R1

Dear Dr. Trogdon,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Ilana Graetz

Academic Editor

PLOS ONE

17 Nov 2021

PONE-D-21-09579R1

Providers’ mediating role for medication adherence among cancer survivors

Dear Dr. Trogdon:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Ilana Graetz

Academic Editor

PLOS ONE