Is a standardized intervention delivered by community health workers effective for improving clinical outcomes across a Veterans Affairs medical center, a federally qualified health center, and an academic family practice clinic?
In this multicenter randomized clinical trial of 592 adults, self-rated health was similar among the intervention and control groups, but the patients in the intervention group were more likely to report high-quality primary care and a reduction in total hospital days.
Interventions by a health system–based community health worker can improve patient-perceived quality of care while reducing hospitalization.
Addressing the social determinants of health has been difficult for health systems to operationalize.
To assess a standardized intervention, Individualized Management for Patient-Centered Targets (IMPaCT), delivered by community health workers (CHWs) across 3 health systems.
Design, Setting, and Participants
This 2-armed, single-blind, multicenter randomized clinical trial recruited patients from 3 primary care facilities in Philadelphia, Pennsylvania, between January 28, 2015, and March 28, 2016. Patients who resided in a high-poverty zip code, were uninsured or publicly insured, and who had a diagnosis for 2 or more chronic diseases were recruited, and patients were randomized to either the CHW intervention or the control arm (goal setting only). Follow-up assessments were conducted at 6 and 9 months after enrollment. Data were analyzed using an intention-to-treat approach from June 2017 to March 2018.
Participants set a chronic disease management goal with their primary care physician; those randomized to the CHW intervention received 6 months of tailored support.
Main Outcomes and Measures
The primary outcome was change in self-rated physical health. The secondary outcomes were self-rated mental health, chronic disease control, patient activation, patient-reported quality of primary care, and all-cause hospitalization.
Of the 592 participants, 370 (62.5%) were female, with a mean (SD) age of 52.6 (11.1) years. Participants in both arms had similar improvements in self-rated physical health (mean [SD], 1.8 [11.2] vs 1.6 [9.9]; P = .89). Patients in the intervention group were more likely to report the highest quality of care (odds ratio [OR], 1.8; 95% CI, 1.4-2.4; risk difference [RD], 0.12; P < .001) and spent fewer total days in the hospital at 6 months (155 days vs 345 days; absolute event rate reduction, 69%) and 9 months (300 days vs 471 days; absolute event rate reduction, 65%). This reduction was driven by a shorter average length of stay (difference, –3.1 days; 95% CI, –6.33 to 0.22; P = .06) and a lower mean number of hospitalizations (difference, –0.3; 95% CI, –0.6 to 0.0; P = .07) among patients who were hospitalized. Patients in the intervention group had a lower odds of repeat hospitalizations (OR, 0.4; 95% CI, 0.2-0.9; RD, –0.24; P = .02), including 30-day readmissions (OR, 0.3; 95% CI, 0.1-0.9; RD, –0.17; P = .04).
Conclusions and Relevance
A standardized intervention did not improve self-rated health but did improve the patient-perceived quality of care while reducing hospitalizations, suggesting that health systems may use a standardized intervention to address the social determinants of health.
ClinicalTrials.gov identifier: NCT02347787
Half of the US population lives with a chronic disease, and 70% of US residents are projected to die from a chronic illness.1-3 The burdens of chronic disease are even greater among people with lower income, who often have multiple chronic conditions3 and face social challenges associated with worse outcomes. More than 20 years ago, sociologists Bruce Link and Jo Phalen4 posited that social challenges (eg, hunger, loneliness, or trauma) were the fundamental causes of a variety of diseases because they placed disadvantaged people at risk of medical risk factors, such as obesity or smoking.
Despite the growing recognition of the importance of social determinants of health, high-quality evidence is scant for effective scalable strategies that health systems may use to address these social determinants.5 Typically, clinicians are unable to address these complex, deeply rooted social issues and instead treat late manifestations of preventable disease.
Community health workers (CHWs) are trusted laypeople from the local community who can be hired and trained by health care organizations to support patients. The CHWs can perform various roles, including providing informal social support, coaching to improve health behaviors, navigating complex health systems, coordinating care, and advocating for patients.6 The CHW interventions are increasingly common and can improve chronic disease outcomes7-12 but are often disease-specific12 or difficult to scale across institutions.13 A growing number of studies have evaluated CHW interventions,6,14 but many of these have been observational or focused exclusively on pediatric populations.
Individualized Management for Patient-Centered Targets (IMPaCT) is a standardized intervention in which CHWs provide tailored social support, navigation, and advocacy to help low-income patients achieve health goals. Two previous single-center randomized clinical trials have studied IMPaCT.15,16 The first trial, involving hospitalized patients, demonstrated that IMPaCT improved access to primary care, mental health, and patient activation and quality while reducing 30-day readmissions.9 The second trial, involving outpatients with multiple chronic conditions, demonstrated that the intervention improved chronic disease control (glycosylated hemoglobin (HbA1c), body mass index (BMI), cigarettes per day (CPD), mental health, and quality of care while reducing hospitalization.6
In this trial, we tested the scalability and effectiveness of IMPaCT in 3 primary care settings important in the care of low-income patients: a Veterans Affairs (VA) medical center, a federally qualified health center, and an academic family practice clinic. We hypothesized that, compared with patients who did not receive CHW support, those enrolled in the IMPaCT program would have improved self-rated health, chronic disease control, patient activation, and quality of primary care as well as lower all-cause hospitalization.
This study was a 2-arm, single-blind (blinded outcomes assessor), parallel-group, multisite randomized clinical trial (NCT02347787) approved by the institutional review boards of the University of Pennsylvania and the Corporal Michael J. Crescenz VA Medical Center. All participants provided written informed consent. The full trial protocol is available in the Supplement.
Study Setting and Participants
Enrollment occurred at a VA primary care practice, a federally qualified health center, and an academic family practice clinic. Eligible participants were (1) adult patients (aged 18 years or older) with an appointment in the previous year; (2) residents of 8 high-poverty zip codes in Philadelphia, Pennsylvania; (3) uninsured or publicly insured, including all veterans; and (4) received a diagnosis for 2 or more targeted chronic diseases (diabetes, obesity, tobacco dependence, and hypertension), at least one of which was in poor control (HbA1c level of 9.0 or higher, BMI of 35 or higher [calculated as weight in kilograms divided by height in meters squared], CPD of more than 0, or systolic blood pressure [SBP] of 160 or greater) (Figure). Patients were excluded if they lacked the capacity to consent.
Each week, trained research assistants (RAs) called eligible patients who had an upcoming appointment to offer study enrollment. Interested patients arrived for their appointments, at which time the RAs obtained their written informed consent, conducted a baseline survey, and recorded their height, weight, blood pressure, and HbA1c level (for patients with a diabetes diagnosis). Patients were recruited between January 28, 2015, and March 28, 2016, at which time the prespecified target sample size was reached.
Patients selected one of their multiple chronic conditions that was in poor control as a focus for the intervention. In consultation with a primary care physician, patients set a chronic disease management goal. The RAs then notified a study team member, who was not involved with outcomes assessment, to perform randomization using a computer-generated algorithm with permuted variable block sizes.17 Randomization was stratified by site and the condition patients selected as their focus. Patients assigned to goal-setting only (control arm) went on to receive usual care. Patients assigned to receive goal-setting plus CHW support (intervention arm) immediately began the 6-month IMPaCT intervention.
The RAs, who were blinded to study arms and hypotheses, conducted an in-person assessment at 6 months after enrollment. To measure the persistence of effect after the intervention ended, the RAs conducted another assessment at 9 months. Assessments included height, weight, blood pressure, and HbA1c. The RAs extracted electronic medical record data, available within approximately 4 weeks after the assessment date, for patients who did not complete follow-up.
IMPaCT15,18-21 occurs in 3 stages: goal-setting, tailored support, and connection with long-term support. At enrollment, CHWs used a semistructured interview guide to get to know the patients holistically and assess their socioeconomic determinants of health (eg, trauma, food insecurity, housing instability, drug and alcohol use, or family stress), as described by Link and Phalen4 and others.22-24 Unlike closed-ended social determinant screeners, the guide used open-ended questions such as, “Some people need a plan for getting enough food … how is that for you?” This approach allowed CHWs to develop a nuanced assessment.
After probing for unmet needs, CHWs asked, “What do you think you’ll need in order to reach the health goal you set with your doctor?” This question allowed patients to express their own goals and preferences. For instance, one patient might consider stable housing and access to fresh vegetables the most important steps toward reaching his health goal, while another patient might describe needing something fun to motivate her to be healthy. These individualized goals became the basis for tailored action plans. The CHWs assessed patient confidence in achieving their goals and advised them to reframe a goal if their confidence was less than 7 on a 10-point scale.
Then, CHWs provided 6 months of hands-on, tailored support spanning the domains of coaching, social support, advocacy, and navigation to help patients achieve their action plans. However, CHWs did not directly provide health education or clinical care, and when these needs arose, CHWs navigated patients to the appropriate clinician. The CHWs communicated with patients at least once per week, including monthly face-to-face contact. If a patient was hospitalized during the intervention, CHWs visited the patient and coordinated with the inpatient care team. For instance, if CHWs determined that the patient could not afford discharge medications or execute instructions, they informed the inpatient care team so that the plans could be modified prior to discharge. After discharge, CHWs helped the patient obtain medications, schedule follow-up appointments, and address other barriers to recovery.
Finally, CHWs helped patients identify long-term supports that could bolster the patients after the intervention ended. The supports may include neighbors, family members, church, or a weekly CHW-facilitated support group.18
Even as the intervention was individualized, leaving room for tailoring in patient care plans, IMPaCT was highly standardized in terms of its approach to hiring, training, workflows, supervision, documentation, and intervention fidelity. Intervention guidelines are codified in the form of detailed manuals, in-person and online training, and software for documentation and reporting.15,18-21 The CHWs are required to have a high school diploma and are screened through behavioral interviews for personality traits, such as empathy. They undergo a month-long training that covers topics such as action planning and motivational interviewing. They are supervised by a manager (typically with a master’s degree in social work) who provides real-time support, ongoing training, and help with clinical integration. Managers reinforce intervention fidelity through weekly assessments: audits of documentation, observation of CHWs in the field, phone calls to patients, and review of performance dashboard.
Prior to this trial, some of us (S.K., T.C., L.N., and J.A.L.) spent 1 to 3 months adapting IMPaCT for each of the 3 sites using a systematic approach previously developed by our full team.20 The approach involves engaging key stakeholders and conducting contextual inquiry to identify areas for which the intervention materials needed to be tailored. For instance, in this study, protocols for how CHWs communicate with primary care teams were tailored on the basis of local preferences. Another example is that the documentation software was rebuilt behind the VA firewall to comply with VA security regulations. Despite these adaptations, the core components of IMPaCT were relatively unchanged, which enabled rapid scale. The intervention was delivered by a separate staff at each institution (including a veteran CHW at the VA). All staff members were trained on site-specific policies and gained work privileges, including access to the medical record and touch-down space.
All outcomes were measured longitudinally over 0 (baseline), 6, and 9 months. The prespecified primary outcome was the mean change in self-rated physical health (using the SF-12v2 Health Survey Physical Component Summary instrument, a 12-item scale measuring physical health components; score range: 0-100, with the highest score indicating maximal physical health) at 6 months.25 This assessment includes questions such as, “Does your health now limit you in moderate activities, such as moving a table, pushing a vacuum cleaner, bowling, or playing golf?” and “Climbing several flights of stairs?” This outcome was chosen on the basis of findings from formative focus groups, where patients prioritized “feeling better physically” over “feeling better emotionally,” “improving chronic disease numbers,” or “staying out of the hospital.”
Prespecified secondary outcomes were the mean change in self-rated mental health (using the SF-12v2 Health Survey Mental Component Summary instrument, a 12-item scale measuring mental health components; score range: 0-100, with the highest score indicating maximal mental health),25 mean change in patient-selected chronic disease (HbA1c, BMI, SBP, or CPD), mean change in patient activation measure,26 proportion of patients reporting high quality of care (the comprehensiveness and support for self-management domains of the Consumer Assessment of Healthcare Providers and Systems Patient-Centered Medical Homes survey),27 and all-cause hospitalization.28 Hospitalization data for nonveterans were collected by linking patient identifiers with the Pennsylvania Health Care Cost Containment Council28 statewide database for all hospitalizations throughout Pennsylvania. Hospitalization data for veterans were assessed using administrative data from the national Veterans Health Administration Corporate Data Warehouse,29 as VA policies did not permit linking identifiers to hospitalization data outside of the VA firewall. After determining that the Pennsylvania Health Care Cost Containment Council could not consistently collect information on emergency department visits, we removed this secondary outcome.
Sample sizes were based on detecting a between-arm difference in mean change in SF-12v2 Health Survey Physical Component Summary at 6 months of 3 points, the minimal clinically significant difference for this instrument.30 To achieve at least 80% power with a type I error rate of 5%, we required 444 total participants. To account for 25% attrition, we aimed to accrue 592 participants. Data were analyzed from June 2017 to March 2018.
Descriptive comparisons between group baseline characteristics were performed with χ2 or Fisher exact tests for categorical variables and with unpaired 2-tailed t tests and the Wilcoxon rank sum test for continuous variables. Event rates for hospitalization were calculated by dividing the total number of hospital days by the number of patients in each arm; the arithmetic difference in these rates by arm was the absolute event rate reduction.
All hypotheses were 2-sided and tested using an intention-to-treat analysis based on random assignment. The primary outcome was measured at 6 months. All secondary outcomes were measured longitudinally over 0, 6 and 9 months. We used generalized estimating equations to account for repeated measures per person over time. Working independence correlation structure with robust empirical variance estimators was used. Appropriate functional models were used: linear regression to estimate the difference in differences for continuous outcomes and31 logistic regression to estimate odds ratios (ORs) for dichotomous outcomes; these models were then converted to risk differences (RDs) for consistency with the difference in differences calculations.32 A stratum-specific multivariate analysis of variance33 was used to measure between-arm difference in change in the patient-selected chronic disease. The multivariate analysis of variance is an extension of analysis of variance, which allowed all patients (regardless of their selected chronic disease focus) to be assessed in a single model.19,33 Hypothesis tests were conducted using joint multivariate Wald test statistics. Permutation-based 2-sided P values were used for these tests to preserve the correct type I error without making any distributional assumptions; P = .05 indicated statistical significance.
Models were adjusted for stratification variables: site and selected chronic disease.34 Adjusted and unadjusted results were similar; adjusted models are presented here for simplicity.
We performed multiple imputation by chained equations for missing data,35 pooling results from 5 imputed data sets according to the Rubin variance formula.36 The imputation model included outcome variables, baseline variables that were imbalanced or stratified at randomization, and variables associated with missing outcome information. As a sensitivity analysis, we conducted a complete-case analysis; the results were nearly identical to those that use imputed data and thus we only report the imputed results.
In total, 1339 patients were screened for eligibility, 42 (3.1%) of whom were ineligible (Figure). Of the remaining 1297 patients, 592 (45.6%) provided consent and were randomized, and 705 (54.4%) declined to participate. Common reasons for declining were lacking the time to enroll (126 [17.9%]), not wanting a CHW (113 [16.0%]), and not wanting to participate in research (103 [14.6%]). Participants, compared with those who declined, were younger (mean age, 52.6 years vs 57.8 years; P < .001) and more likely to be female (372 [62.6%] vs 322 [45.9%]; P < .001). Of the 592 patients, 370 (62.5%) were female and 222 (37.5%) were male, with a mean (SD) age of 52.6 (11.1) years. The race/ethnicity of most patients (558 [94.3%]) was African American, and 577 patients (97.8%) had experienced a traumatic event37 (Table 1).
Complete primary outcome data were available in nearly equal numbers in both study arms at 6 months (222 [77.1%] vs 248 [81.6%]; P = .18) and at 9 months (227 [78.8%] vs 233 [76.6%]; P = .53). Patients who did not complete both study arms were more likely to be younger (50.0 years vs 53.3 years; P = .003), have lower self-rated mental health (40.4 vs 43.2; P = .04), report delayed or unmet health needs (67 [54.9%] vs 205 [43.6%]; P = .02 and 42 [34.4] vs 115 [24.5%]; P = .03), list the emergency department as their usual care setting (13 [1.7%] vs 25 [5.4%]; P = .04), have a more anxious attachment style (mean [SD] score, 3.2 [1.8] vs 2.4 [1.7]; P = .03; an 11-item scale from 1-6, with the highest score indicating the greatest proclivity to that attachment style), and have a lower health literacy mean (SD) score (2.4 [1.5] vs 1.9 [1.2]; P = .05; score range: 1-5, with the highest score indicating low literacy). Admissions data were available for 100% of participants as the data were obtained from claims.
Participants had a mean (SD) number of 2.5 (0.6) chronic conditions, and most patients selected tobacco dependence or obesity as their focus for the intervention. Patients were in poor control of their selected conditions; the mean (SD) baseline values for each condition were as follows: diabetes, HbA1c level of 10.5% (2.0%); obesity, BMI of 42.5 (7.5); tobacco dependence, CPD of 9.3 (6.8); and hypertension, SBP of 162 (23.3) mm Hg. The 2 study arms were similar in all baseline characteristics with the exception of Hispanic ethnicity (0 vs 11 [3.7%]; P < .001).
At 6 months, the intervention arm had less improvement in self-rated physical health compared with the control arm (mean [SD], 0.6 [12.7] vs 2.3 [11.3]; P = .06). However, by 9 months, both groups had similar improvement in self-rated physical health (mean [SD], 1.8 [11.2] vs 1.6 [9.9]; P = .89). Longitudinal estimates, taking into account both the 6- and 9-month measures, confirmed that both arms had similar improvements in self-rated physical health (difference in differences, –0.7; 95% CI, –2.2 to 0.7; P = .30) and self-rated mental health (difference in differences, 0.8; 95% CI, –1.1 to 2.6; P = .41) (Table 2). The intervention arm had a greater improvement in patient activation (difference in differences, 1.9; 95% CI, –0.1 to 3.8; P = .06). Intervention patients were more likely to report the highest quality of care pertaining to comprehensiveness (OR, 1.8; 95% CI, 1.4-2.4; RD, 0.12; P < .001) and disease self-management (OR, 1.8; 95% CI, 1.4-2.4; RD, 0.12; P < .001) (Table 3). The intervention arm had similar improvements in chronic disease control (difference in differences: HbA1c, –0.23%; BMI, –0.15; CPD, –0.50; SBP, –6.26 mm Hg [P = .21]).
Patients in the intervention arm compared with patients in the control arm spent fewer total days in the hospital at 6 months (155 days vs 345 days; absolute event rate reduction, 69%) and 9 months (300 days vs 471 days; absolute event rate reduction, 65%). These differences were explained by a shorter average length of stay (difference, –3.1 days; 95% CI, –6.33 to 0.22; P = .06) and a lower mean number of hospitalizations (difference, –0.3; 95% CI, –0.6 to 0.0; P = .07) among patients who were hospitalized. Specifically, intervention patients who were hospitalized were less likely to need readmission: They had a lower odds of repeat hospitalizations (OR, 0.4; 95% CI, 0.2-0.9; RD, –0.24; P = .02), including 30-day readmissions (OR, 0.3; 95% CI, 0.1-0.9; RD, –0.17; P = .04).
This trial was not powered to detect differences by site; however, the main study findings appeared consistent across sites, and no statistically significant interactions were observed between arm and site.
Overall, 277 of 304 patients (91.0%) engaged with the CHW intervention for the full 6 months. The remaining 27 patients (8.9%) were lost to the CHW despite 10 outreach attempts (including 1 home visit) or said they no longer wanted to work with a CHW. Patients and CHWs created a mean (SD) of 5.5 (2.0) action plans together, most commonly in the categories of psychosocial issues (524 [36.9%]) and health behavior (505 [35.6%]). Example activities that CHWs engaged in to help participants with their action plans included among others: attending social events; exercising with participants; helping participants apply for social services; helping participants link to appropriate specialist care; telling providers when a participant could not afford a prescribed medication so a lower cost alternative could be identified; and linking participants to community based mental health care and drug addiction treatment. Patients completed 60.3% of their action plans. No statistically significant differences in process or outcome measures were found among the CHWs delivering the intervention.
To improve outcomes among low-income, chronically ill populations, health systems must reach beyond their walls and address socioeconomic and behavioral factors. This randomized clinical trial tests IMPaCT, a standardized, health system-based social CHW intervention across multiple diseases and settings.5 We found that the intervention did not improve self-rated health but did lead to substantial and sustained improvements in perceived quality of primary care and reductions in hospital use. These findings are consistent with 2 previous trials of the IMPaCT intervention, which demonstrated modest improvements in patient activation and chronic disease control as well as larger effects on patient-reported quality of care and hospitalization.
This current study adds to the body of evidence supporting the role of CHWs as trusted mediators between disadvantaged patients and formal health care organizations.38-40 Although primary care practices employ other personnel (eg, social workers, nurses, and physicians) who play critical roles in the clinical team, they are unlikely to share a background with disadvantaged patients. Empirical tests of dynamic social impact theory41,42 and social network and comparison theories43 support the notion that this shared identity allows CHWs to establish trust,21 offer practical support based on their own experience of common problems,44,45 and provide nonjudgmental support.44,46
The intervention and control groups achieved similar improvement in the primary outcome, self-rated physical health. It is possible that the intervention did not affect this aspect of health. Another possible explanation is that intervention patients became more active (eg, exercising at the YMCA with their CHWs), calling attention to physical limitations of which they would have otherwise been unaware. A qualitative process evaluation conducted concurrently with this trial will help us to explore these questions.
The two largest effects of the CHW intervention were perceived quality of primary care and hospitalizations, which may be related. Previous studies47,48 have suggested that low-income patients may perceive hospital-based care as more accessible and of higher quality than the primary care available to them. This perception of imbalance can result in patients preferring to go to the hospital instead of the doctor’s office. The CHWs in this study were closely integrated with primary care practices and served as liaisons between clinicians and patients, which appears to have improved patients’ overall perception of their primary care. These improvements are in and of themselves important to health care organizations, given the reporting and associated incentives for measures of the patient experience. In addition, this increased sense that the primary care practice was delivering comprehensive care that was supportive of self-management may have caused patients to rely more on the clinicians rather than the hospital for acute exacerbations of chronic disease. The support that CHWs provided if patients were hospitalized (eg, communicating with the inpatient care team, helping with postdischarge transition) may also have had a direct effect on both length of stay and repeat hospitalization. Support from CHWs may have led inpatient care teams to feel comfortable discharging patients more quickly. Furthermore, CHWs may have also helped prevent repeat admissions by facilitating rapid primary care follow-up16 and guiding patients to address social or financial obstacles known to precipitate rapid return, such as difficulty paying for discharge medications.
Several factors make IMPaCT well-suited for practical use by health systems. Community-based interventions are often viewed as setting-specific and slow to replicate. A strength of this intervention was its standardized approach to hiring, training, supervision, caseloads, and workflow. The intervention even standardized the approach to tailoring the program for individual patients and new clinical sites. The resulting program fit the needs of each patient and site but was still generalizable across different institutions with minimal variation in outcomes. In addition, benefits were seen over a relatively short time horizon and treatment effects persisted for at least the 3 months we observed after the study period. Taken together, these findings suggest that health systems could see rapid and potentially lasting benefits from a single scalable CHW intervention.
This trial has limitations. First, all sites were located in Philadelphia; future studies will test the scalability of the intervention across different locations. Second, we do not know if the effects persisted beyond the 9 months of the trial. Third, hospitalization data for veterans were limited to the VA network. Fourth, some data were missing, although relatively little considering the vulnerable patient population. Finally, as with all patient-level trials, external validity can be limited because participants differ from those who decline. However, to our knowledge, this trial’s enrollment rates were higher than those in many published sociobehavioral intervention studies.5,49
This study supports the notion that health care organizations can use a standardized CHW intervention to address socioeconomic and behavioral factors and improve quality of care while reducing hospitalization.
Accepted for Publication: July 6, 2018.
Corresponding Author: Shreya Kangovi, MD, MS, University of Pennsylvania, 423 Guardian Dr, 1233 Blockley Hall, Philadelphia, PA 19104 (email@example.com).
Published Online: October 22, 2018. doi:10.1001/jamainternmed.2018.4630
Author Contributions: Dr Kangovi had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Kangovi, Norton, Harte, Grande, Long.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Kangovi, Norton, Long.
Critical revision of the manuscript for important intellectual content: Mitra, Norton, Carter, Harte, Zhao, Grande, Long.
Statistical analysis: Kangovi, Mitra, Zhao, Long.
Obtained funding: Kangovi, Long.
Administrative, technical, or material support: Norton, Carter, Harte, Zhao, Grande.
Supervision: Kangovi, Norton, Long.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was funded in part by grant K23-HL128837 from the National Institutes of Health National Heart, Lung, and Blood Institute (Dr Kangovi) and by grant PCORI-1310-07292 from the Patient-Centered Outcomes Research Institute (all authors). Research reported in this publication was funded through a Patient-Centered Outcomes Research Institute Award.
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Meeting Presentation: This paper was presented as a plenary session of the Society of General Internal Medicine National Meeting; April 13, 2018; Denver, Colorado.
Disclaimer: The statements in this publication are solely the responsibility of the authors and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute, its Board of Governors, or Methodology Committee.
Centers for Medicare & Medicaid Services. Chronic Conditions Among Medicare Beneficiaries: 2012 Edition. Chart Book. Baltimore, MD: Centers for Medicare & Medicaid Services;2012.
et al. The Seattle-King County healthy homes project: implementation of a comprehensive approach to improving indoor environmental quality for low-income children with asthma. Environ Health Perspect
. 2002;110(suppl 2):311-322. doi:10.1289/ehp.02110s2311PubMedGoogle ScholarCrossref
et al. Effects of community-based health worker interventions to improve chronic disease management and care among vulnerable populations: a systematic review. Am J Public Health
. 2016;106(4):e3-e28. doi:10.2105/AJPH.2015.302987PubMedGoogle ScholarCrossref
JA. Community health worker support for disadvantaged patients with multiple chronic diseases: a randomized clinical trial. Am J Public Health
. 2017;107(10):1660-1667. doi:10.2105/AJPH.2017.303985PubMedGoogle ScholarCrossref
Randomization BS (for Windows) [computer program]. Version 6.0. Los Angeles, CA; 2010.
et al. Penn Center for community health workers. http://chw.upenn.edu/
. Accessed September 18, 2018.
JA. A randomized controlled trial of a community health worker intervention in a population of patients with multiple chronic diseases: study design and protocol. Contemp Clin Trials
. 2017;53:115-121. doi:10.1016/j.cct.2016.12.009PubMedGoogle ScholarCrossref
et al. Toward a scalable, patient-centered community health worker model: adapting the IMPaCT intervention for use in the outpatient setting. Popul Health Manag
. 2016;19(6):380-388. doi:10.1089/pop.2015.0157PubMedGoogle ScholarCrossref
AW. The Institute of Medicine report “Unequal Treatment”: implications for academic health centers. Mt Sinai J Med
. 2004;71(5):314-321.PubMedGoogle Scholar
M. A Guide to the Development of Certified Short Form Interpretation and Reporting Capabilities. Lincoln, RI: Quality Metric Inc; 2009.
Department of Veterans Affairs. 172VA10P2: VHA Corporate Data Warehouse – VA. 79 FR 4377. Fed Regist
. 2014;79(17):4377-4382.Google Scholar
DM. User’s Manual for the SF-12v2 Health Survey. Lincoln, RI: Quality Metric; 2010.
DB. Statistical Analysis With Missing Data
. 2nd ed. Hoboken, NJ: Wiley; 2002, doi:10.1002/9781119013563
L. Community health workers can be a public health force for change in the United States: three actions for a new paradigm. Am J Public Health
. 2011;101(12):2199-2203. doi:10.2105/AJPH.2011.300386PubMedGoogle ScholarCrossref