Abstract
Objectives. To test the effect of CommunityRx, a scalable, low-intensity intervention that matches patients to community resources, on mental health-related quality of life (HRQOL) (primary outcome), physical HRQOL, and confidence in finding resources.
Methods. A real-world trial assigned publicly insured residents of Chicago, Illinois, aged 45 to 74 years to an intervention (n = 209) or control (n = 202) group by alternating calendar week, December 2015 to August 2016. Intervention group participants received usual care and an electronic medical record–generated, personalized list of community resources. Surveys (baseline, 1-week, 1- and 3-months) measured HRQOL and confidence in finding community resources to manage health.
Results. At 3 months, there was no difference between groups in mental (–1.03; 95% confidence interval [CI] = −3.02, 0.96) or physical HRQOL (0.59; 95% CI = −0.98, 2.16). Confidence in finding resources was higher in the intervention group (odds ratio = 2.08; 95% CI = 1.18, 3.63); the effect increased at each successive time point. Among intervention group participants, 65% recalled receiving the intervention; 48% shared community resource information with others.
Conclusions. CommunityRx did not increase HRQOL, but its positive effect on confidence in finding resources for self-care suggests that this low-intensity intervention may have a role in population health promotion.
Trial Registration. ClinicalTrials.gov Identifier: NCT02435511.
The shift from fee-for-service to value-based payment models in health care demands population health management strategies that support individuals’ efforts to maintain health and manage chronic conditions outside the health care setting.1 Policy recommendations for population health management have called for cross-sector collaboration, including hand-offs from medical professionals to community-based service providers2,3 that support individuals’ self-care, basic, or “health-related social needs.”4 Although community referral solutions are already being adopted by health systems, little is known about the effect of these interventions on, or the mechanisms through which they might improve, health outcomes.5–7
CommunityRx is a community resource referral information system, developed in partnership with stakeholders across sectors, local residents, and with support from a US Center for Medicare and Medicaid Innovation Healthcare Innovation Award.4 The delivery approach is “whole person,”8 universal, and low-intensity. By making meaningful use of electronic medical record (EMR) data9 and integrating with existing clinical workflows, CommunityRx addresses the full range of resource needs for all people seeking health care. At each visit, for every person, a HealtheRx is generated, including resources for basic needs such as food and housing, physical and mental wellness, and disease management including smoking cessation, weight loss, and counseling. Data about community resources, including location, hours, and cost, are obtained by ongoing assessment via direct observation and phone surveys.10,11
In a previous observational study of CommunityRx, more than 113 000 outpatients (aged 0–99 years) received a HealtheRx during a clinical encounter (Appendix A, available as a supplement to the online version of this article at http://www.ajph.org). Each HealtheRx listed community resources matched to the patient’s characteristics (age, gender, preferred language, home address) and conditions (e.g., wellness, homelessness, Alzheimer’s disease, hypertension). This study demonstrated the feasibility of integrating the automated CommunityRx intervention with widely used EMR systems and yielded positive process-level outcomes including feasibility and high satisfaction among patients and providers.12
The Center for Medicare and Medicaid Innovation also commissioned an external evaluation of all Health Care Innovation Awards, including CommunityRx, to assess their impact on health care utilization. This case–control study found, during the 3 years following implementation, an average increase in primary care visits and decrease in hospital admissions among Medicare beneficiaries for whom at least 1 HealtheRx was generated (n = 7385) compared with 7260 matched controls and an average decrease in emergency department (ED) visits among Medicaid beneficiaries (n = 2408) compared with 2437 matched controls.13,14 Although these early results are promising, they do not provide insights into the individual-level mechanism of action or the effect of the CommunityRx intervention on health outcomes.
To characterize the individual-level mechanisms and health effects of CommunityRx, we adapted the Self- and Family Management Framework by Grey et al.15 widely used to study interventions to promote chronic disease management (Appendix B, available as a supplement to the online version of this article at http://www.ajph.org). The framework identifies evidence-based “processes” or targets for interventions, including “activating community resources” as a key process that drives more distal quality of life, health, and health care outcomes.15 CommunityRx targets this key process. The model also differentiates between proximal and distal outcomes, identifying self-efficacy as 1 of several important proximal outcomes on the causal pathway linking self-management to improved health.
Other promising resource referral interventions have been delivered to more selected populations, often using a screening tool, targeted to a single basic or social need (e.g., food insecurity or intimate partner violence) or specific disease (e.g., diabetes, stroke) and either manually created or not personally tailored.7 Most of these interventions have been studied under efficacy conditions,16 delivered by trained research personnel, rather than via the usual workflow. To fill a gap in knowledge about the effectiveness of a universal, low-intensity approach, we conducted a real-world trial17 of CommunityRx with an inclusive sample of middle-aged and older adults. Building on the framework by Grey et al., we tested the hypotheses that providing automatically generated referrals to community resources for all patients at the point of care would yield near-term improvement in patient mental health-related quality of life (HRQOL; primary outcome) and physical HRQOL (secondary outcome). We also tested the hypothesis that the intervention would increase confidence in finding resources for self-care, a measure of self-efficacy (secondary outcome).
METHODS
This trial was conducted at an urban academic medical center serving a predominantly African American (54% of 961 000 people) and high-poverty region (51% of families lived below 200% of the federal poverty level).18 Results reporting follows the Transparent Reporting of Evaluations with Nonrandomized Designs guidelines.19
Study Participants
Participants were enrolled from December 2015 to August 2016 from the ED and primary care clinic (PCC). The last follow-up survey was conducted in December 2016. Patients aged 45 to 74 years seeking care were eligible if they were beneficiaries of Medicare, Medicaid, or both, and resided in the 16-zip-code study region (inclusive of the medical center’s primary service area). Those who recalled previously receiving a HealtheRx, did not speak English, or lacked capacity to provide informed consent because of cognitive status or medical acuity were ineligible.
Study Procedures
The automated HealtheRx intervention was integrated with routine EMR-based discharge workflows. Delivery of the intervention by clinical staff via their usual workflow prohibited individual-level randomization; patients were assigned by alternating calendar week to the intervention or control groups.
Survey instruments were pretested for validity and feasibility of administration. Research assistants conducted interviews by using Qualtrics (Qualtrics, Provo, UT). Face-to-face baseline interviews were conducted during the medical encounter. Phone interviews were conducted at 1 week, 1 month, and 3 months. The baseline survey assessed sociodemographic characteristics and outcome measures; follow-up surveys included outcome measures and, for the intervention group only, questions about the HealtheRx. Participants received a $15 gift card for baseline survey completion and a $25 check payment for each completed follow-up survey.
Intervention
Details of the CommunityRx system have been described previously.12 Control group participants received usual care, which could include receiving oral or written information about resources. In addition to usual care, intervention group participants received the HealtheRx from their nurse in the ED or administrative staff in PCC during discharge. Intervention group participants also received a mailed copy of their HealtheRx with the 1-week interview incentive check.
To ensure that intervention group participants received the HealtheRx and control group participants did not, research staff in the PCC observed participants’ discharge process. Observation was not logistically possible for participants enrolled in the ED; instead, signs reminding ED staff of the protocol were posted. Three months into the trial, a digital “on/off” switch was added to the CommunityRx software to manage alternating week assignment and, at the patient level, to prevent a HealtheRx from being generated for control group patients who returned for scheduled care during their 3-month follow-up. Nonetheless, some control participants (e.g., patients who had an unscheduled or same-day scheduled visit) may have received 1 or more HealtheRxs during the 3-month follow-up.
Intervention group participants who returned to the ED or PCC during their follow-up period could receive additional HealtheRxs. Medical records were queried to estimate the total number of HealtheRxs potentially received by every participant. For controls, any ED or PCC visit during the 3-month follow-up period for which there was no record of intercepting the HealtheRx was counted as a potential exposure.
Outcome Measures
The primary outcome was mental HRQOL measured using the Short Form-12 (SF-12) mental composite score (MCS, range = 0–100).20 Secondary outcomes included (1) physical HRQOL measured using the SF-12 physical composite score (PCS; range = 0–100)20 and (2) confidence in finding resources measured using an item developed from Bandura’s self-efficacy theory21: “How confident are you in your ability to find resources in your community that help you manage your health?” (not at all, not very, uncertain, somewhat, or completely). This item was constructed for this study as there was no established measure for assessing confidence in finding resources.
To assess fidelity, we assessed intervention group participants’ recall of receiving the HealtheRx and resource use. To estimate the rate of community resource use attributable to usual care, control group participants who said they received resource information were asked “Since you received this information, have you gone to any of the places?” Intervention group participants who recalled receiving the HealtheRx were asked the same question, but told to think specifically about places on the HealtheRx. Given the differences in administration of these questions, intervention and control group responses cannot be compared.
Statistical Analysis
The primary analysis compared mean MCS between intervention and control groups using the intent-to-treat principle. A mixed-effects linear regression model22 was fit to each continuous outcome, including treatment group (intervention or control), time point (as a discrete factor), and their interaction, permitting a different effect at each time point. Additional covariates included baseline score, intervention location (PCC or ED), age (decades), gender, education, race, and ethnicity to account for possible differences between groups attributable to nonrandom treatment assignment and to increase precision of estimates. We used a patient-level random intercept to account for within-person correlation among the 3 follow-up measurements, and we used the robust (i.e., sandwich) variance estimator in case of a violation of the exchangeable correlation assumption.23 We fit the model by using maximum likelihood. We used a similar mixed-effects ordinal logistic regression to analyze confidence in finding resources.
For each model, we calculated estimated differences between treatment groups, both separately at each time point and overall. Higher MCS and PCS indicate better mental and physical HRQOL, respectively. We used Wald tests to test the null hypothesis of no effect, and we calculated 2-sided P values and corresponding 95% confidence intervals (CIs). Although we used all available outcome data in each model (Appendix C, available as a supplement to the online version of this article at http://www.ajph.org), we excluded 1 participant with missing MCS and PCS data at baseline as well as 34 (8%) individuals with missing education values.
To examine possible bias attributable to missing data, we reestimated the models by using multiple imputation. The pattern of missing data was not monotone (even among the outcomes), and education and confidence in finding resources were not continuous; we therefore used Multiple Imputation using Chained Equations.23 We imputed MCS and PCS with linear regression, we imputed confidence in finding resources with ordinal logistic regression, and we imputed education with multinomial logistic regression. We used all covariates (including intervention group) and outcomes in each imputation model (e.g., we used each timepoint in imputing the others, and we used MCS, PCS, and confidence in finding resources in each imputation model). We used a burn-in period of 100 iterations, and we checked convergence by using trace plots. We generated 100 sets of imputations.
A power calculation performed at the time of study design assumed a 2-sample comparison of means and a standard deviation for MCS of 10 (10% of the score range). A sample of 400 (200 per group) provided 80% power to detect a difference in mean MCS of 2.8, assuming a 2-sided test at the .05 level. A 3-point difference in HRQOL scores in both cross-sectional and longitudinal comparisons (12–24 months) is considered clinically meaningful.24 This calculation was conservative as it did not include adjustment for covariates, which improves the model’s precision.
As described previously, both control and intervention group participants may have been exposed to 1 or more HealtheRxs. Although one might hypothesize that the effect of the intervention would be proportional to the number of HealtheRxs received, a naïve analysis examining the association between number of potential exposures and the outcomes would yield a biased estimate of the intervention effect if patients with poorer health have more frequent return visits (and therefore more potential exposures). To address this, we used an endogenous treatment model,25 which permits estimation of the effect of exposure to the HealtheRx on each outcome assuming group assignment is exogenous (i.e., unrelated to patients’ outcomes except through its effect on the number of HealtheRxs received). We fit this model to each outcome at 3 months, using the total number of HealtheRxs potentially received as a covariate and adjusting for the baseline value of the outcome, location where the intervention was received, age, gender, education, race, and ethnicity (Appendix D, available as a supplement to the online version of this article at http://www.ajph.org).
Using data from the intervention group only, we fit mixed-effects logistic regression models to test the association between selected covariates and the (1) likelihood of recalling receipt of the HealtheRx and (2) likelihood of telling someone else about the HealtheRx, conditional on having recalled receiving it. The covariates included time point (as a discrete factor) and demographic variables; we also included a random intercept for each patient.
We conducted analyses with Stata version 15.1 (StataCorp LP, College Station, TX).
RESULTS
A total of 420 people were enrolled, the majority being African American (90%) and female (68%); 411 completed the baseline interview and were included in the analysis (209 intervention, 202 control; Figure 1). Eighty-five percent of participants completed the 1-week follow-up and 80% completed the 1- and 3-month follow-ups; 9% of participants completed no follow-up interviews (Appendix C). Follow-up completion rates were similar for both groups at all time points, but consistently higher (9–17 percentage points) among those enrolled in the PCC compared with the ED.
FIGURE 1—
Trial Enrollment Participants Included in Intent-to-Treat Analysis: Chicago, IL, December 2015–August 2016
Groups were similar at baseline (Table 1), indicating that group assignment by alternating calendar week achieved good balance. At baseline, the rate of poor or fair health was high (51%), the overall mean physical HRQOL was low (mean = 39; range = 14–62). Two participants, 1 in each group, died during the follow-up period. Sixty-five percent of intervention group participants recalled receiving a HealtheRx.
TABLE 1—
Study Participant Characteristics: Chicago, IL, December 2015–August 2016
| Intervention (n = 209) | Control (n = 202) | |
| Age, y, no. (%) | ||
| 45–54 | 54 (25.8) | 62 (30.7) |
| 55–64 | 76 (36.4) | 65 (32.2) |
| 65–74 | 79 (37.8) | 75 (37.1) |
| Gender, no. (%) | ||
| Female | 152 (72.7) | 129 (63.9) |
| Male | 57 (27.3) | 73 (36.1) |
| Race/ethnicity, no. (%) | ||
| Non-Hispanic Black | 187 (89.5) | 183 (90.6) |
| Non-Hispanic White | 11 (5.3) | 12 (5.9) |
| Hispanic/Latino | 3 (1.4) | 3 (1.5) |
| Other or ≥ 1 race | 8 (3.8) | 4 (2.0) |
| Education, no. (%) | ||
| < high school | 39 (18.7) | 52 (25.7) |
| High school or GED | 43 (20.6) | 37 (18.3) |
| Associate degree or some college | 78 (37.3) | 61 (30.2) |
| College degree | 34 (16.3) | 33 (16.3) |
| Annual household income, $, no. (%) | ||
| < 25 000 | 101 (48.3) | 118 (56.5) |
| 25 000–49 999 | 62 (29.7) | 44 (21.1) |
| 50 000–99 999 | 8 (3.8) | 9 (4.3) |
| ≥ 100 000 | 8 (3.8) | 6 (2.9) |
| Self-reported health, no. (%) | ||
| Poor or fair | 104 (49.8) | 104 (51.5) |
| Good | 68 (32.5) | 60 (29.7) |
| Very good or excellent | 37 (17.7) | 38 (18.8) |
| Confidence in finding resources, no. (%) | ||
| Not at all confident | 22 (10.5) | 25 (12.4) |
| Not very confident | 22 (10.5) | 26 (12.9) |
| Uncertain | 27 (12.9) | 23 (11.4) |
| Somewhat confident | 70 (33.5) | 62 (30.7) |
| Completely confident | 68 (32.5) | 66 (32.7) |
| Baseline SF-12, no. (mean) | ||
| Mental health component score | 208 (48.7) | 202 (50.0) |
| Physical health component score | 208 (38.8) | 202 (38.6) |
Note. GED = general educational development; SF-12 = Short Form-12. The sample size was n = 411. Missing data not shown. Percentage may not equal 100 because of rounding.
Mean MCS scores did not change over time among participants in either group (48.7 at baseline vs 48.7, 49.2, and 48.3 at 1 week, 1 month, and 3 months among the intervention group; 50.0 at baseline vs 49.6, 50.2, and 50.0 among the control group). Differences between treatment groups in mean MCS during the follow-up period, conditional on baseline MCS and the other covariates, were not significantly different from zero (Table 2, Appendix E, available as a supplement to the online version of this article at http://www.ajph.org). Mean PCS decreased slightly in both groups at 1 week, but then returned to baseline levels at 1 and 3 months (38.8 at baseline vs 36.7, 38.2, and 38.7 among the intervention group; 38.6 vs 37.3, 37.7, and 38.1 among the control group). However, the differences between groups in mean PCS during the follow-up period were also not significantly different from zero (Appendix F, available as a supplement to the online version of this article at http://www.ajph.org).
TABLE 2—
Estimated Intervention Effects (Intervention Group Minus Control Group) on Mean Short Form-12 Mental Component Score and Physical Component Score at 1 Week, 1 Month, and 3 Months: Chicago, IL, December 2015–December 2016
| Model 1, b (95% CI) | Model 2, b (95% CI) | |
| Mental component score | ||
| 1 wk | 0.25 (−1.65, 2.15) | 0.18 (−1.73, 2.08) |
| 1 mo | −0.07 (−2.01, 1.87) | −0.17 (−2.10, 1.76) |
| 3 mo | −0.98 (−2.99, 1.02) | −1.03 (−3.02, 0.96) |
| Overall | −0.27 (−1.82, 1.29) | −0.34 (−1.89, 1.21) |
| Physical component score | ||
| 1 wk | −0.58 (−2.07, 0.91) | −0.34 (−1.83, 1.16) |
| 1 mo | 0.25 (−1.32, 1.82) | 0.46 (−1.10, 2.02) |
| 3 mo | 0.40 (−1.18, 1.99) | 0.59 (−0.98, 2.16) |
| Overall | 0.02 (−1.17, 1.22) | 0.24 (−0.95, 1.43) |
Note. CI = confidence interval. Based on mixed effects linear regression models including data from all completed follow-up time points. Model 1 was adjusted for baseline score only. Model 2 was adjusted for baseline mental or physical component score, location where the intervention was received (primary care clinic or emergency department), age (decades), gender, education, race, and ethnicity.
Confidence in finding resources increased among both groups relative to baseline; however, the increase among the intervention group was greater. The percentage of intervention group participants answering “somewhat confident” or “completely confident” increased from 66% at baseline to 69%, 73%, and 80% at 1 week, 1 month, and 3 months; among the control group the percentages increased from 63% at baseline to 68%, 67%, and 69% at each follow-up point. At 3 months, the odds of reporting a higher category of confidence among the intervention group were double those for the control group (odds ratio [OR] = 2.1; 95% CI = 1.2, 3.6), with adjustment for baseline confidence and covariates (Table 3, Appendix G, available as a supplement to the online version of this article at http://www.ajph.org). Re-estimating this model with multiple imputation yielded a similar result (OR = 1.56; 95% CI = 1.05, 2.34; see Appendices H, I, and J, available as supplements to the online version of this article at http://www.ajph.org, for results with multiple imputation).
TABLE 3—
Estimated Intervention Effect on Confidence in Finding Resources: Chicago, IL, December 2015 to December 2016
| Outcome | Model 1,a OR (95% CI) | Model 2,b AOR (95% CI) | Model 3,c AOR (95% CI) |
| Comparing the intervention group to the control group | |||
| Confidence in finding resources | |||
| 1 wk | 1.38 (0.79, 2.42) | 1.37 (0.78, 2.40) | |
| 1 mo | 1.63 (0.93, 2.88) | 1.63 (0.92, 2.89) | |
| 3 mo | 2.07 (1.18, 3.64) | 2.07 (1.18, 3.63) | |
| Overall | 1.67 (1.09, 2.57) | 1.67 (1.08, 2.56) | |
| Endogenous treatment model estimating effect of the number of HealtheRxs received | |||
| Confidence in finding resources: 3 mo | 1.09 (0.95, 1.25) | 1.31 (1.03, 1.66) | |
Note. AOR = adjusted odds ratio; CI = confidence interval; OR = odds ratio. Effects estimated by using mixed-effects ordinal regression. Model 1 includes adjustment for baseline confidence in finding resources only; models 2 and 3 include adjustment for baseline confidence in finding resources and location where intervention was received (primary care clinic or emergency department), age (decades), gender, education, race, and ethnicity.
Adjusted for baseline score only.
Adjusted for covariates.
Adjusted for covariates, accounting for endogeneity.
Among 174 intervention group participants with 3-month outcome data, 61% received 1 or 2 HealtheRxs (1 at baseline and 1 mailed home after the 1-week interview), 22% received 3 HealtheRxs, and 16% received more than 3. Among 158 control group participants with 3-month outcome data, 20% had 1 visit during the 3-month follow-up period during which they may have received a HealtheRx, 9% had 2 visits, and 8% had 3 or more visits (Appendix K, available as a supplement to the online version of this article at http://www.ajph.org). The number of visits with possible receipt of a HealtheRx was not significantly associated with MCS or PCS, with adjustment for baseline score and the covariates (Appendices L and M, available as supplements to the online version of this article at http://www.ajph.org). By contrast, each additional visit was associated with an increase in confidence in finding resources (OR = 1.3; 95% CI = 1.0, 1.7; Table 3; Appendix N, available as a supplement to the online version of this article at http://www.ajph.org).
The rate of community resource use resulting from usual care, estimated from the control group, was 17%. Of all participants in the intervention group, 11% reported using a community resource listed on the HealtheRx; among the 65% of participants in the intervention group who recalled receiving a HealtheRx, 14% reported using a community resource listed on the HealtheRx, and 48% reported sharing the information from the HealtheRx with others. The likelihood of recall and information sharing increased over the follow-up period (Appendix O, available as a supplement to the online version of this article at http://www.ajph.org). Intervention group participants who, at baseline, reported being “not at all confident” about finding community resources to manage their health were less likely than others to recall receiving a HealtheRx; other individual characteristics were not associated with recall.
DISCUSSION
This trial evaluated the impact of the automated CommunityRx intervention, delivered via routine workflows with minimal training, on health outcomes for a clinically heterogeneous population of middle-aged and older adults presenting for primary and low-acuity emergency care. We found no significant effect of this intervention on mental or physical HRQOL during the 3-month follow-up period. This follow-up interval may have been too short to realize the intervention’s health benefits for middle-aged and older adults with a high burden of disease.
In contrast, we found a significant effect of the intervention on individuals’ confidence in finding community resources to manage their health. We assessed confidence by using a measure informed by Bandura’s self-efficacy theory21 and posited that increased confidence (“strength of belief”) to find community resources (“ability to execute a requisite action”) would drive more distal health outcomes by promoting healthful community resource use.26 This positive finding on a proximal outcome corroborates our underlying theory of behavior change.15 Although our domain-specific measure is mechanistically more proximal than general self-efficacy,27 self-efficacy is emerging as a key factor on the causal pathway linking illness management processes to more distal health outcomes.15,28–30 The CommunityRx intervention targets this process by routinely providing all patients high-quality, personalized information from a trusted source about relevant community resources.
The optimal strategy to assess impact of a community resource referral intervention on use of community resources would be to obtain individual-level use data from community-based service providers. Similar to previous evaluations of interventions that referred patients to multiple resource types,31,32 CommunityRx relied on participants’ self-report of resource use. Because of the low-intensity intervention and real-world design, we assumed that not everyone in the intervention group would receive or recall receiving the intervention. Among those who recalled receiving the HealtheRx intervention, 14% reported using a community resource they found on the HealtheRx. This rate does not account for resource use resulting from usual care alone (17% among controls). This finding is similar to resource use rates reported for a volunteer sample (n = 458) in the previous CommunityRx observational study12 and for higher-intensity interventions delivered by specially trained clinicians or staff.31,33 To our knowledge, this is the first study to report a rate of community resource use resulting from an automated intervention delivered under real-world conditions.
Unlike drug and device interventions, information-based interventions delivered in a medical setting can easily spread beyond the individual patient. In this study, 48% of intervention group participants who recalled receiving the intervention shared information from the HealtheRx with others. This finding replicates evidence from the previous CommunityRx study, which found a 49% sharing rate. To our knowledge, studies of other community resource referral interventions have not queried this important dynamic effect. This finding has 2 important implications. First, studies of community resource referral interventions may undervalue the impact on resource use and more distal outcomes if limited to individual-level analysis. Second, a universal, rather than screening-based, approach to community resource referrals may have greater potential for spread. With more than 113 000 participants in CommunityRx studies to date, no negative or adverse consequences of the universal delivery approach have surfaced. Because CommunityRx is an automated, EMR-integrated solution that requires little training to deliver, the per-capita cost of a universal delivery approach is marginal.
Limitations
We acknowledge limitations that should be considered when interpreting the results. Without random assignment, confounding may bias the estimated treatment effects. However, the 2 groups were well-balanced, and several relevant variables were included in the models. In addition, the real-world design may have contributed heterogeneity to the observed treatment effect; specifically, the possible exposure of some control participants to HealtheRx may have reduced our estimated treatment effect. Accordingly, our endogenous treatment models account for the highest possible number of HealtheRxs received by each participant and yield similar results to the intent-to-treat analyses. Community resource use rates among intervention group participants could not be meaningfully compared with use rates reported by control group participants. Lastly, this study does not yield insight on the quality or relative health value of resource types used by participants.
Public Health Implications
Changes in payment policy are incentivizing health systems to adopt practices that optimize individual and population health. Evidence-based and scalable interventions that serve whole people, rather than individual diseases, are urgently needed to promote efforts to prevent and manage chronic disease outside of the clinical care system. In alignment with the science of behavior change theory, sustainable adoption of community resource referral interventions in the real world will require that public health and health care professionals understand how and why these interventions work.16
ACKNOWLEDGMENTS
Research reported in this publication was supported by the National Institute on Aging of the National Institutes of Health R01AG047869 (S. T. L., PI). The full amount of the project costs were financed with federal money.
Note: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
CONFLICTS OF INTEREST
S. T. Lindau directed a Center for Medicare and Medicaid Innovation Health Care Innovation Award (1C1CMS330997-03) called CommunityRx. This award required development of a sustainable business model to support the model test after award funding ended. To this end, S. T. Lindau is founder and co-owner of NowPow LLC and President of MAPSCorps, 501c3. Neither The University of Chicago nor The University of Chicago Medicine is endorsing or promoting any NowPow or MAPSCorps entity or its business, products, or services. S. T. Lindau and her spouse own stocks and mutual funds managed by a third company. S. T. Lindau and her spouse are shareholders in Glenbervie Health LLC, a company with no relationship to the topic of this study.
HUMAN PARTICIPANT PROTECTION
This study was approved the by The University of Chicago institutional review board. All participants provided written informed consent.
Footnotes
REFERENCES
- 1.Burwell SM. Setting value-based payment goals—HHS efforts to improve U.S. health care. N Engl J Med. 2015;372(10):897–899. doi: 10.1056/NEJMp1500445. [DOI] [PubMed] [Google Scholar]
- 2.Edwards ST, Landon BE. Medicare’s chronic care management payment—payment reform for primary care. N Engl J Med. 2014;371(22):2049–2051. doi: 10.1056/NEJMp1410790. [DOI] [PubMed] [Google Scholar]
- 3.Weinstein JN, Geller A, Neguisse Y, Baciu A. Communities in Action: Pathways to Health Equity. Washington, DC: National Academies Press; 2017. Policies to support community solutions. In: National Academies of Sciences, Engineering, and Medicine; pp. 335–382. [PubMed] [Google Scholar]
- 4.Alley DE, Asomugha CN, Conway PH, Sanghavi DM. Accountable health communities—addressing social needs through Medicare and Medicaid. N Engl J Med. 2016;374(1):8–11. doi: 10.1056/NEJMp1512532. [DOI] [PubMed] [Google Scholar]
- 5.Fierman AH, Beck AF, Chung EK et al. Redesigning health care practices to address childhood poverty. Acad Pediatr. 2016;16(3 suppl):S136–S146. doi: 10.1016/j.acap.2016.01.004. [DOI] [PubMed] [Google Scholar]
- 6.Gottlieb LM, Tirozzi KJ, Manchanda R, Burns AR, Sandel MT. Moving electronic medical records upstream: incorporating social determinants of health. Am J Prev Med. 2015;48(2):215–218. doi: 10.1016/j.amepre.2014.07.009. [DOI] [PubMed] [Google Scholar]
- 7.Gottlieb LM, Wing H, Adler NE. A systematic review of interventions on patients’ social and economic needs. Am J Prev Med. 2017;53(5):719–729. doi: 10.1016/j.amepre.2017.05.011. [DOI] [PubMed] [Google Scholar]
- 8.Shim R, Rust G. Primary care, behavioral health, and public health: partners in reducing mental health stigma. Am J Public Health. 2013;103(5):774–776. doi: 10.2105/AJPH.2013.301214. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Centers for Medicare and Medicaid Services. Medicare & Medicaid EHR Incentive Program: meaningful use stage 1 requirements overview 2010. Available at: https://www.cms.gov/Regulations-and-Guidance/Legislation/EHRIncentivePrograms/downloads/MU_Stage1_ReqOverview.pdf. Accessed September 21, 2018.
- 10.Makelarski JA, Lindau ST, Fabbre VD et al. Are your asset data as good as you think? Conducting a comprehensive census of built assets to improve urban population health. J Urban Health. 2013;90(4):586–601. doi: 10.1007/s11524-012-9764-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Lindau ST, James R, Makelarski JA, Sanders E, Johnson D. Comments from the south side of Chicago on New Haven’s Inspiring Initiative. Am J Public Health. 2012;102(7):e3–34; author reply e4–35. doi: 10.2105/AJPH.2012.300684. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Lindau ST, Makelarski J, Abramsohn E et al. CommunityRx: a population health improvement innovation that connects clinics to communities. Health Aff (Millwood) 2016;35(11):2020–2029. doi: 10.1377/hlthaff.2016.0694. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.RTI International. Evaluation of the Health Care Innovation Awards: Community Resource Planning, Prevention and Monitoring, Third Annual Report. 2017. Available at: https://downloads.cms.gov/files/cmmi/hcia-communityrppm-thirdannualrpt.pdf. Accessed April 4, 2018.
- 14.RTI International. Evaluation of the Health Care Innovation Awards: Community Resource Planning, Prevention, and Monitoring. Third Annual Report Addendum. 2017. Available at: https://downloads.cms.gov/files/cmmi/hcia-crppm-thirdannrptaddendum.pdf. Accessed April 16, 2018.
- 15.Grey M, Schulman-Green D, Knafl K, Reynolds NR. A revised self- and family management framework. Nurs Outlook. 2015;63(2):162–170. doi: 10.1016/j.outlook.2014.10.003. [DOI] [PubMed] [Google Scholar]
- 16.Onken LS, Carroll KM, Shoham V, Cuthbert BN, Riddle M. Reenvisioning clinical science: unifying the discipline to improve the public health. Clin Psychol Sci. 2014;2(1):22–34. doi: 10.1177/2167702613497932. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Ford I, Norrie J. Pragmatic trials. N Engl J Med. 2016;375(5):454–463. doi: 10.1056/NEJMra1510059. [DOI] [PubMed] [Google Scholar]
- 18.US Census Bureau. 2016. American Community Survey 5-year estimates. 2016. Available at: http://factfinder.census.gov. Accessed June 2, 2018.
- 19.Des Jarlais DC, Lyles C, Crepaz N. TREND Group. Improving the reporting quality of nonrandomized evaluations of behavioral and public health interventions: the TREND statement. Am J Public Health. 2004;94(3):361–366. doi: 10.2105/ajph.94.3.361. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Ware J, Kosinski M, Keller SD. A 12-Item Short-Form Health Survey: construction of scales and preliminary tests of reliability and validity. Med Care. 1996;34(3):220–233. doi: 10.1097/00005650-199603000-00003. [DOI] [PubMed] [Google Scholar]
- 21.Bandura A. Self Efficacy: The Exercise of Control. New York, NY: WH Freeman; 1997. [Google Scholar]
- 22.McCulloch CE, Searle SR, Neuhaus JM. Generalized, Linear, and Mixed Models. Hoboken, NJ: John Wiley & Sons; 2008. [Google Scholar]
- 23.White IR, Royston P, Wood AM. Multiple imputation using chained equations: issues and guidance for practice. Stat Med. 2011;30(4):377–399. doi: 10.1002/sim.4067. [DOI] [PubMed] [Google Scholar]
- 24.Samsa G, Edelman D, Rothman ML, Williams GR, Lipscomb J, Matchar D. Determining clinically important differences in health status measures: a general approach with illustration to the Health Utilities Index Mark II. Pharmacoeconomics. 1999;15(2):141–155. doi: 10.2165/00019053-199915020-00003. [DOI] [PubMed] [Google Scholar]
- 25.Kenkel DS, Terza JV. The effect of physician advice on alcohol consumption: count regression with an endogenous treatment effect. J Appl Econ. 2001;16(2):165–184. [Google Scholar]
- 26.Bandura A. Guide for constructing self-efficacy scales. In: Self-Efficacy Beliefs of Adolescents. Information Age Publishing. 2006:307–337. Available at: https://www.uky.edu/∼eushe2/Bandura/BanduraGuide2006.pdf. Accessed February 23, 2018.
- 27.Bandura A. Self-efficacy: toward a unifying theory of behavioral change. Psychol Rev. 1977;84(2):191–215. doi: 10.1037//0033-295x.84.2.191. [DOI] [PubMed] [Google Scholar]
- 28.Churchill SS, Kieckhefer GM. One year follow-up of outcomes from the randomized clinical trial of the Building on Family Strengths Program. Matern Child Health J. 2018;22(6):913–921. doi: 10.1007/s10995-018-2467-4. [DOI] [PubMed] [Google Scholar]
- 29.Schulman-Green D, Jaser S, Martin F et al. Processes of self-management in chronic illness. J Nurs Scholarsh. 2012;44(2):136–144. doi: 10.1111/j.1547-5069.2012.01444.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Ryan P, Sawin KJ. The individual and family self-management theory: background and perspectives on context, process, and outcomes. Nurs Outlook. 2009;57(4):217–225.e6. doi: 10.1016/j.outlook.2008.10.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Garg A, Toy S, Tripodis Y, Silverstein M, Freeman E. Addressing social determinants of health at well child care visits: a cluster RCT. Pediatrics. 2015;135(2):e296–e304. doi: 10.1542/peds.2014-2888. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Tremont G, Davis JD, Ott BR et al. Randomized trial of the family intervention: telephone tracking-caregiver for dementia caregivers: use of community and healthcare resources. J Am Geriatr Soc. 2017;65(5):924–930. doi: 10.1111/jgs.14684. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Smith S, Malinak D, Chang J et al. Implementation of a food insecurity screening and referral program in student-run free clinics in San Diego, California. Prev Med Rep. 2016;5:134–139. doi: 10.1016/j.pmedr.2016.12.007. [DOI] [PMC free article] [PubMed] [Google Scholar]