Identifying and characterizing cancer survivors in the US primary care safety net

Megan Hoopes; Teresa Schmidt; Nathalie Huguet; Kerri Winters-Stone; Heather Angier; Miguel Marino; Jackilen Shannon; Jennifer DeVoe

doi:10.1002/cncr.32295

Identifying and characterizing cancer survivors in the US primary care safety net

Megan Hoopes, Teresa Schmidt, Nathalie Huguet, Kerri Winters-Stone, Heather Angier, Miguel Marino, Jackilen Shannon, Jennifer DeVoe

Research output: Contribution to journal › Article › peer-review

12 Scopus citations

Abstract

Background: Primary care providers must understand the use patterns, clinical complexity, and primary care needs of cancer survivors to provide quality health care services. However, to the authors' knowledge, little is known regarding the prevalence and health care needs of this growing population, particularly in safety net settings. Methods: The authors identified adults with a history of cancer documented in primary care electronic health records within a network of community health centers (CHCs) in 19 states. The authors estimated cancer history prevalence among >1.2 million patients and compared sex-specific site distributions with national estimates. Each survivor was matched to 3 patients without cancer from the same set of clinics. The demographic characteristics, primary care use, and comorbidity burden then were compared between the 2 groups, assessing differences with absolute standardized mean differences (ASMDs). ASMD values >0.1 denote meaningful differences between groups. Generalized estimating equations yielded adjusted odds ratios (aORs) for select indicators. Results: A total of 40,266 cancer survivors were identified (prevalence of 3.0% of adult CHC patients). Compared with matched cancer-free patients, a higher percentage of survivors had ≥6 primary care visits across 3 years (62% vs 48%) and were insured (83% vs 74%) (ASMD, >0.1 for both). Cancer survivors had excess medical complexity, including a higher prevalence of depression, asthma/chronic obstructive pulmonary disease, and liver disease (ASMD, >0.1 for all). Survivors had higher odds of any opioid prescription (aOR, 1.23; 95% CI, 1.19-1.27) and chronic opioid therapy (aOR, 1.27; 95% CI, 1.23-1.32) compared with matched controls (P <.001 for all). Conclusions: Identifying cancer survivors and understanding their patterns of utilization and physical and mental comorbidities present an opportunity to tailor primary health care services to this population.

Original language	English (US)
Pages (from-to)	3448-3456
Number of pages	9
Journal	Cancer
Volume	125
Issue number	19
DOIs	https://doi.org/10.1002/cncr.32295
State	Published - Oct 1 2019

Keywords

electronic health records
neoplasms
primary health care
survivorship

ASJC Scopus subject areas

Oncology
Cancer Research

Access to Document

10.1002/cncr.32295

Cite this

@article{b38200e974334af28400f23a0ee5dab5,

title = "Identifying and characterizing cancer survivors in the US primary care safety net",

abstract = "Background: Primary care providers must understand the use patterns, clinical complexity, and primary care needs of cancer survivors to provide quality health care services. However, to the authors' knowledge, little is known regarding the prevalence and health care needs of this growing population, particularly in safety net settings. Methods: The authors identified adults with a history of cancer documented in primary care electronic health records within a network of community health centers (CHCs) in 19 states. The authors estimated cancer history prevalence among >1.2 million patients and compared sex-specific site distributions with national estimates. Each survivor was matched to 3 patients without cancer from the same set of clinics. The demographic characteristics, primary care use, and comorbidity burden then were compared between the 2 groups, assessing differences with absolute standardized mean differences (ASMDs). ASMD values >0.1 denote meaningful differences between groups. Generalized estimating equations yielded adjusted odds ratios (aORs) for select indicators. Results: A total of 40,266 cancer survivors were identified (prevalence of 3.0% of adult CHC patients). Compared with matched cancer-free patients, a higher percentage of survivors had ≥6 primary care visits across 3 years (62% vs 48%) and were insured (83% vs 74%) (ASMD, >0.1 for both). Cancer survivors had excess medical complexity, including a higher prevalence of depression, asthma/chronic obstructive pulmonary disease, and liver disease (ASMD, >0.1 for all). Survivors had higher odds of any opioid prescription (aOR, 1.23; 95% CI, 1.19-1.27) and chronic opioid therapy (aOR, 1.27; 95% CI, 1.23-1.32) compared with matched controls (P <.001 for all). Conclusions: Identifying cancer survivors and understanding their patterns of utilization and physical and mental comorbidities present an opportunity to tailor primary health care services to this population.",

keywords = "electronic health records, neoplasms, primary health care, survivorship",

author = "Megan Hoopes and Teresa Schmidt and Nathalie Huguet and Kerri Winters-Stone and Heather Angier and Miguel Marino and Jackilen Shannon and Jennifer DeVoe",

note = "Funding Information: Jennifer DeVoe received a grant from the National Cancer Institute for work performed as part of the current study. The other authors made no disclosures. Supported by grant 1R01CA204267-01 from the National Cancer Institute at the National Institutes of Health. The research reported in this article was conducted with the Accelerating Data Value Across a National Community Health Center Network (ADVANCE) Clinical Research Network, a partner of the National Patient-Centered Clinical Research Network (PCORnet), an initiative originally funded by the Patient-Centered Outcomes Research Institute (PCORI) and now funded by the People-Centered Research Foundation (PCRF). The ADVANCE network is led by OCHIN in partnership with the Health Choice Network, Fenway Health, Oregon Health and Science University, and the Robert Graham Center and HealthLandscape. ADVANCE is funded through PCRF contract number 1237. The Accelerating Data Value Across a National Community Health Center Network (ADVANCE) clinical data research network, one of 13 PCORnet (an initiative originally funded by the Patient-Centered Outcomes Research Institute [PCORI] and now funded by the People-Centered Research Foundation [PCRF]) clinical data research networks, is a multicenter collaborative led by the OCHIN (formerly the Oregon Community Health Information Network but shortened to OCHIN when it was expanded to other states) network in partnership with the Health Choice Network (based in Florida) and Fenway Health (based in Massachusetts). The 3 member organizations are comprised of CHCs spanning 24 states that serve disadvantaged and vulnerable populations. EHR data from each partner organization are integrated and standardized into a common data model. The demographic profile of the ADVANCE patient population mirrors that of national CHC estimates and is generalizable to the US safety net population. The Accelerating Data Value Across a National Community Health Center Network (ADVANCE) clinical data research network, one of 13 PCORnet (an initiative originally funded by the Patient-Centered Outcomes Research Institute [PCORI] and now funded by the People-Centered Research Foundation [PCRF]) clinical data research networks, is a multicenter collaborative led by the OCHIN (formerly the Oregon Community Health Information Network but shortened to OCHIN when it was expanded to other states) network in partnership with the Health Choice Network (based in Florida) and Fenway Health (based in Massachusetts). The 3 member organizations are comprised of CHCs spanning 24 states that serve disadvantaged and vulnerable populations. EHR data from each partner organization are integrated and standardized into a common data model. The demographic profile of the ADVANCE patient population mirrors that of national CHC estimates and is generalizable to the US safety net population. We identified 42,359 cancer survivors who were alive and aged ≥19 years as of December 31, 2016. To be included, patients had to have ≥1 office visits at 1 of 431 primary care clinics that were actively using an EHR system throughout the 3-year study period (2014-2016), and had to have a record in a discrete searchable EHR field (ie, encounter diagnosis, problem list, medical history) indicating a malignant cancer diagnosis, excluding nonmelanoma skin cancer. These cancer survivors represented approximately 3.0% of the CHC patient population in the current study within the 19 states with active primary care clinics in the ADVANCE network. To construct a comparison group, we identified >1.2 million adult patients from the same clinics with no documented cancer diagnosis as of December 31, 2016. This population was younger and had a different sex distribution compared with cancer survivors and thus to reduce bias and improve group comparisons, cancer survivors were matched in a 1:3 ratio to this comparison group based on sex, year of birth, and primary health system (each patient's most frequently accessed CHC). For each cancer survivor, 3 matches were selected at random and without replacement. This exact matching method resulted in the exclusion of 2093 cancer survivors because they had <3 control matches available (4.9% of survivors). Excluded patients were more often male, non-Hispanic white, and older than the patients with cancer who were included in the current study. The final data set contained 40,226 cancer survivors with 47,339 primary cancer sites (because a single patient can have >1 cancer site) and 120,798 matched comparison patients. Cancer diagnoses were grouped into primary sites following classifications from the US Surveillance, Epidemiology, and End Results (SEER) program (https://seer.cancer.gov/tools/conversion/). To provide estimates of data completeness and general population representativeness, we compared sex-stratified rankings and prevalence for the most common cancer sites with national estimates. For each survivor, we calculated the number of distinct cancer sites recorded, length of time since diagnosis, and age at the time of the first cancer diagnosis (based on EHR-reported onset or noted dates), and compared this information with national estimates when available (see Supporting Tables 1 and 2). Cancer survivors were compared with matched cancer-free counterparts based on race, ethnicity, preferred language, and a household income ≤138% of the Federal Poverty Level (FPL; the cutpoint for Medicaid eligibility under the Patient Protection and Affordable Care Act). We also compared the groups based on insurance type, length of time established at their primary CHC, whether patients were ever uninsured or consistently uninsured throughout the study period, number of office visits, number of providers seen, and primary care visits. We compared cancer survivors with their matched cancer-free counterparts based on the Charlson Comorbidity Index (CCI), pain, and opioid use. An enhanced version of the CCI, which considers additional physical, mental, and behavioral health conditions, was applied to the EHR problem list. We disregarded cancer diagnoses from our CCI calculation for a more accurate comparison of noncancer comorbidity burden between the groups. The 10 most prevalent conditions among the population in the current study were presented individually. Chronic pain diagnoses were identified from problem list records, and prescription orders were used to identify those individuals who were prescribed opioid medications. Chronic opioid use was defined as ≥60 opioid tablets or any fentanyl patch prescribed within any 90-day period. We identified opioid use disorder using both problem list and encounter diagnoses. All time-varying characteristics (eg, FPL, CCI) were assigned as of each patient's last encounter during the study period. The current cross-sectional analysis used data across a 3-year study period (2014-2016). In univariable analyses, we presented patient demographics, health care use, and the prevalence of chronic conditions by cancer history groups. Given the large sample size, we assessed differences between the groups using absolute standardized mean differences (ASMDs), which are not affected by sample size. The ASMD is an effect size measure that is increasingly used in observational studies to compare distributional differences between groups, with extensions to binary and multinomial variables. The ASMD is defined as the difference in group means in units of SD, and ranges from 0 (indicating that the groups are equivalent on the measure being compared) to 1.0 (indicating perfect disagreement). As has been done in other studies, we considered an ASMD of >0.1 (interpreted as a 10% difference) to denote meaningful differences between the groups. In multivariable analyses, we compared 4 utilization indicators for cancer survivors compared with their cancer-free counterparts controlling for important confounders through regression modeling. To account for the correlated data arising from the relatedness of cancer survivors and their cancer-free counterparts through matching, we used generalized estimating equation logistic regression models, with SEs clustered on match set to test for differences between cancer survivors and their cancer-free counterparts. We computed adjusted odds ratios (aORs) with 95% CIs adjusted for race/ethnicity, CCI, insurance type, years established, PCP assignment, and number of office visits. Statistical significance was 2-sided and set at α =.05. Due to the clinical heterogeneity of cancer and its treatment, we also have presented demographic, utilization, and clinical measures for survivors stratified by the leading 5 cancer sites for men and women (see Supporting Table 3). Data management and analysis were conducted using SAS statistical software (version 9.4; SAS Institute Inc, Cary, North Carolina). The institutional review board of the Oregon Health and Science University approved the current study. Publisher Copyright: {\textcopyright} 2019 American Cancer Society",

year = "2019",

month = oct,

day = "1",

doi = "10.1002/cncr.32295",

language = "English (US)",

volume = "125",

pages = "3448--3456",

journal = "Cancer",

issn = "0008-543X",

publisher = "John Wiley and Sons Inc.",

number = "19",

}

TY - JOUR

T1 - Identifying and characterizing cancer survivors in the US primary care safety net

AU - Hoopes, Megan

AU - Schmidt, Teresa

AU - Huguet, Nathalie

AU - Winters-Stone, Kerri

AU - Angier, Heather

AU - Marino, Miguel

AU - Shannon, Jackilen

AU - DeVoe, Jennifer

N1 - Funding Information: Jennifer DeVoe received a grant from the National Cancer Institute for work performed as part of the current study. The other authors made no disclosures. Supported by grant 1R01CA204267-01 from the National Cancer Institute at the National Institutes of Health. The research reported in this article was conducted with the Accelerating Data Value Across a National Community Health Center Network (ADVANCE) Clinical Research Network, a partner of the National Patient-Centered Clinical Research Network (PCORnet), an initiative originally funded by the Patient-Centered Outcomes Research Institute (PCORI) and now funded by the People-Centered Research Foundation (PCRF). The ADVANCE network is led by OCHIN in partnership with the Health Choice Network, Fenway Health, Oregon Health and Science University, and the Robert Graham Center and HealthLandscape. ADVANCE is funded through PCRF contract number 1237. The Accelerating Data Value Across a National Community Health Center Network (ADVANCE) clinical data research network, one of 13 PCORnet (an initiative originally funded by the Patient-Centered Outcomes Research Institute [PCORI] and now funded by the People-Centered Research Foundation [PCRF]) clinical data research networks, is a multicenter collaborative led by the OCHIN (formerly the Oregon Community Health Information Network but shortened to OCHIN when it was expanded to other states) network in partnership with the Health Choice Network (based in Florida) and Fenway Health (based in Massachusetts). The 3 member organizations are comprised of CHCs spanning 24 states that serve disadvantaged and vulnerable populations. EHR data from each partner organization are integrated and standardized into a common data model. The demographic profile of the ADVANCE patient population mirrors that of national CHC estimates and is generalizable to the US safety net population. The Accelerating Data Value Across a National Community Health Center Network (ADVANCE) clinical data research network, one of 13 PCORnet (an initiative originally funded by the Patient-Centered Outcomes Research Institute [PCORI] and now funded by the People-Centered Research Foundation [PCRF]) clinical data research networks, is a multicenter collaborative led by the OCHIN (formerly the Oregon Community Health Information Network but shortened to OCHIN when it was expanded to other states) network in partnership with the Health Choice Network (based in Florida) and Fenway Health (based in Massachusetts). The 3 member organizations are comprised of CHCs spanning 24 states that serve disadvantaged and vulnerable populations. EHR data from each partner organization are integrated and standardized into a common data model. The demographic profile of the ADVANCE patient population mirrors that of national CHC estimates and is generalizable to the US safety net population. We identified 42,359 cancer survivors who were alive and aged ≥19 years as of December 31, 2016. To be included, patients had to have ≥1 office visits at 1 of 431 primary care clinics that were actively using an EHR system throughout the 3-year study period (2014-2016), and had to have a record in a discrete searchable EHR field (ie, encounter diagnosis, problem list, medical history) indicating a malignant cancer diagnosis, excluding nonmelanoma skin cancer. These cancer survivors represented approximately 3.0% of the CHC patient population in the current study within the 19 states with active primary care clinics in the ADVANCE network. To construct a comparison group, we identified >1.2 million adult patients from the same clinics with no documented cancer diagnosis as of December 31, 2016. This population was younger and had a different sex distribution compared with cancer survivors and thus to reduce bias and improve group comparisons, cancer survivors were matched in a 1:3 ratio to this comparison group based on sex, year of birth, and primary health system (each patient's most frequently accessed CHC). For each cancer survivor, 3 matches were selected at random and without replacement. This exact matching method resulted in the exclusion of 2093 cancer survivors because they had <3 control matches available (4.9% of survivors). Excluded patients were more often male, non-Hispanic white, and older than the patients with cancer who were included in the current study. The final data set contained 40,226 cancer survivors with 47,339 primary cancer sites (because a single patient can have >1 cancer site) and 120,798 matched comparison patients. Cancer diagnoses were grouped into primary sites following classifications from the US Surveillance, Epidemiology, and End Results (SEER) program (https://seer.cancer.gov/tools/conversion/). To provide estimates of data completeness and general population representativeness, we compared sex-stratified rankings and prevalence for the most common cancer sites with national estimates. For each survivor, we calculated the number of distinct cancer sites recorded, length of time since diagnosis, and age at the time of the first cancer diagnosis (based on EHR-reported onset or noted dates), and compared this information with national estimates when available (see Supporting Tables 1 and 2). Cancer survivors were compared with matched cancer-free counterparts based on race, ethnicity, preferred language, and a household income ≤138% of the Federal Poverty Level (FPL; the cutpoint for Medicaid eligibility under the Patient Protection and Affordable Care Act). We also compared the groups based on insurance type, length of time established at their primary CHC, whether patients were ever uninsured or consistently uninsured throughout the study period, number of office visits, number of providers seen, and primary care visits. We compared cancer survivors with their matched cancer-free counterparts based on the Charlson Comorbidity Index (CCI), pain, and opioid use. An enhanced version of the CCI, which considers additional physical, mental, and behavioral health conditions, was applied to the EHR problem list. We disregarded cancer diagnoses from our CCI calculation for a more accurate comparison of noncancer comorbidity burden between the groups. The 10 most prevalent conditions among the population in the current study were presented individually. Chronic pain diagnoses were identified from problem list records, and prescription orders were used to identify those individuals who were prescribed opioid medications. Chronic opioid use was defined as ≥60 opioid tablets or any fentanyl patch prescribed within any 90-day period. We identified opioid use disorder using both problem list and encounter diagnoses. All time-varying characteristics (eg, FPL, CCI) were assigned as of each patient's last encounter during the study period. The current cross-sectional analysis used data across a 3-year study period (2014-2016). In univariable analyses, we presented patient demographics, health care use, and the prevalence of chronic conditions by cancer history groups. Given the large sample size, we assessed differences between the groups using absolute standardized mean differences (ASMDs), which are not affected by sample size. The ASMD is an effect size measure that is increasingly used in observational studies to compare distributional differences between groups, with extensions to binary and multinomial variables. The ASMD is defined as the difference in group means in units of SD, and ranges from 0 (indicating that the groups are equivalent on the measure being compared) to 1.0 (indicating perfect disagreement). As has been done in other studies, we considered an ASMD of >0.1 (interpreted as a 10% difference) to denote meaningful differences between the groups. In multivariable analyses, we compared 4 utilization indicators for cancer survivors compared with their cancer-free counterparts controlling for important confounders through regression modeling. To account for the correlated data arising from the relatedness of cancer survivors and their cancer-free counterparts through matching, we used generalized estimating equation logistic regression models, with SEs clustered on match set to test for differences between cancer survivors and their cancer-free counterparts. We computed adjusted odds ratios (aORs) with 95% CIs adjusted for race/ethnicity, CCI, insurance type, years established, PCP assignment, and number of office visits. Statistical significance was 2-sided and set at α =.05. Due to the clinical heterogeneity of cancer and its treatment, we also have presented demographic, utilization, and clinical measures for survivors stratified by the leading 5 cancer sites for men and women (see Supporting Table 3). Data management and analysis were conducted using SAS statistical software (version 9.4; SAS Institute Inc, Cary, North Carolina). The institutional review board of the Oregon Health and Science University approved the current study. Publisher Copyright: © 2019 American Cancer Society

PY - 2019/10/1

Y1 - 2019/10/1

N2 - Background: Primary care providers must understand the use patterns, clinical complexity, and primary care needs of cancer survivors to provide quality health care services. However, to the authors' knowledge, little is known regarding the prevalence and health care needs of this growing population, particularly in safety net settings. Methods: The authors identified adults with a history of cancer documented in primary care electronic health records within a network of community health centers (CHCs) in 19 states. The authors estimated cancer history prevalence among >1.2 million patients and compared sex-specific site distributions with national estimates. Each survivor was matched to 3 patients without cancer from the same set of clinics. The demographic characteristics, primary care use, and comorbidity burden then were compared between the 2 groups, assessing differences with absolute standardized mean differences (ASMDs). ASMD values >0.1 denote meaningful differences between groups. Generalized estimating equations yielded adjusted odds ratios (aORs) for select indicators. Results: A total of 40,266 cancer survivors were identified (prevalence of 3.0% of adult CHC patients). Compared with matched cancer-free patients, a higher percentage of survivors had ≥6 primary care visits across 3 years (62% vs 48%) and were insured (83% vs 74%) (ASMD, >0.1 for both). Cancer survivors had excess medical complexity, including a higher prevalence of depression, asthma/chronic obstructive pulmonary disease, and liver disease (ASMD, >0.1 for all). Survivors had higher odds of any opioid prescription (aOR, 1.23; 95% CI, 1.19-1.27) and chronic opioid therapy (aOR, 1.27; 95% CI, 1.23-1.32) compared with matched controls (P <.001 for all). Conclusions: Identifying cancer survivors and understanding their patterns of utilization and physical and mental comorbidities present an opportunity to tailor primary health care services to this population.

AB - Background: Primary care providers must understand the use patterns, clinical complexity, and primary care needs of cancer survivors to provide quality health care services. However, to the authors' knowledge, little is known regarding the prevalence and health care needs of this growing population, particularly in safety net settings. Methods: The authors identified adults with a history of cancer documented in primary care electronic health records within a network of community health centers (CHCs) in 19 states. The authors estimated cancer history prevalence among >1.2 million patients and compared sex-specific site distributions with national estimates. Each survivor was matched to 3 patients without cancer from the same set of clinics. The demographic characteristics, primary care use, and comorbidity burden then were compared between the 2 groups, assessing differences with absolute standardized mean differences (ASMDs). ASMD values >0.1 denote meaningful differences between groups. Generalized estimating equations yielded adjusted odds ratios (aORs) for select indicators. Results: A total of 40,266 cancer survivors were identified (prevalence of 3.0% of adult CHC patients). Compared with matched cancer-free patients, a higher percentage of survivors had ≥6 primary care visits across 3 years (62% vs 48%) and were insured (83% vs 74%) (ASMD, >0.1 for both). Cancer survivors had excess medical complexity, including a higher prevalence of depression, asthma/chronic obstructive pulmonary disease, and liver disease (ASMD, >0.1 for all). Survivors had higher odds of any opioid prescription (aOR, 1.23; 95% CI, 1.19-1.27) and chronic opioid therapy (aOR, 1.27; 95% CI, 1.23-1.32) compared with matched controls (P <.001 for all). Conclusions: Identifying cancer survivors and understanding their patterns of utilization and physical and mental comorbidities present an opportunity to tailor primary health care services to this population.

KW - electronic health records

KW - neoplasms

KW - primary health care

KW - survivorship

UR - http://www.scopus.com/inward/record.url?scp=85067414231&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85067414231&partnerID=8YFLogxK

U2 - 10.1002/cncr.32295

DO - 10.1002/cncr.32295

M3 - Article

C2 - 31174231

AN - SCOPUS:85067414231

SN - 0008-543X

VL - 125

SP - 3448

EP - 3456

JO - Cancer

JF - Cancer

IS - 19

ER -

Identifying and characterizing cancer survivors in the US primary care safety net

Abstract

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this