Integrated Care for Chronic Conditions: A Randomized Care Management Trial

ABSTRACT

Objectives: Health care systems employ community-based solutions to help individuals manage multiple chronic conditions (MCC). Little knowledge exists on how to optimally translate evidence-based integrated care management models into widespread improvements in patient-centered outcomes. This study aimed to compare effectiveness and differential impact of 3 integrated care management delivery methods.

Study Design: Individual, stratified randomized trial with a 2:2:1 ratio for high-touch (in person), high-tech (remote monitoring), and optimal discharge planning (ODP; telephonic) delivery methods with 12-month follow-up.

Methods: The UPMC Health Plan provided care management to adult Medicaid and Medicare-Medicaid beneficiaries with MCC who were recently discharged from an inpatient hospitalization. Primary (90-day readmission, health status, patient activation) and secondary (30-day readmission; functional status; quality of life; care satisfaction; emergent care use; engagement in primary, specialty, and mental health care; gaps in care) outcomes were assessed.

Results: The analytic sample (n = 1387) included Medicaid (79.5%) or dually eligible (20.5%) beneficiaries with MCC (63.0% female, 73.0% White, and 21.6% Black). We found no evidence of a treatment effect on 90-day readmission rates (P = .669). There was a significant improvement over time for health status (P < .0001) but no significant difference by intervention (P = .866). Patient activation showed a significant time-by-treatment interaction (P = .021), with a significant difference (from baseline to 12 months) for the high-touch approach compared with ODP (adjusted difference of 2.69 points; SE = 1.22; P = .028). Participant subgroups (race, age, illness complexity, and comorbid behavioral health conditions) showed no statistically significant differences by interventions or over time on the primary outcomes.

Conclusions: Results demonstrate how health care systems can leverage a variety of impactful, person-centered care management approaches without compromising patient outcomes.

Am J Manag Care. 2026;32(4):In Press

Takeaway Points

We examined 3 ways of delivering payer-led care management (in person, remote monitoring, telephonic) and found a clinically meaningful change in patient activation over 1 year for those receiving in-person vs telephonic care management. No difference was found between approaches for health status, 90-day readmissions, or subgroups. These findings can:

• inform decision-making by payers and health care systems around varying care management approaches (ie, in person, remote monitoring, telephonic) to increase efficiencies and reach without compromising outcomes;
• guide equitable service implementation to best suit member preferences, staffing structures, and digital resources; and
• extend literature by providing outcomes for interventions that address needs of members with multiple chronic conditions as opposed to one condition.

More than a quarter of US adults are living with multiple chronic conditions (MCC).¹ The growing aging population has resulted in an increase in both prevalence of MCC and demand for best management.¹ Individuals with comorbid behavioral and physical health conditions, those with low income, and those from underserved communities experience especially high hospitalization rates and poor health outcomes.^2-4 Fragmented health care and social service infrastructure, conflicting provider recommendations, and complex social needs contribute to adverse outcomes for individuals with MCC,^5-7 but care coordination and self-management programs can improve outcomes.^8-10 Although chronic condition care models exist,¹¹ health care systems continue to explore methods to implement and scale care management interventions to best support those with MCC.^6,12 In light of the current aging and shrinking of the health care workforce,¹³ health care systems may benefit from having more information about how to adapt effective programs to most successfully meet the needs of Medicaid and Medicaid-Medicare (dually eligible) beneficiaries with complex health and social needs.¹⁴

We sought to compare the effectiveness of 3 integrated care management delivery methods (high-touch, high-tech, and optimal discharge planning [ODP]) on several primary (ie, 90-day hospital readmission, health status, patient activation) and secondary (ie, 30-day hospital readmission; functional status; quality of life; care satisfaction; emergent care use; engagement in primary, specialty, and mental health care; gaps in care) outcomes (aim 1). We hypothesized that the high-touch approach would result in lower readmissions at 12 months compared with the high-tech and ODP approaches (hypothesis [H] 1a); that the high-touch approach would result in higher health status at 12 months vs high-tech and ODP (H1b); and that the high-tech approach would result in higher patient activation at 12 months compared with the high-touch and ODP approaches (H1c). We also examined the differential effects of the interventions for patient subgroups (aim 2). A future paper will discuss the third aim: examining the perceived barriers and facilitators to efficient and effective implementation.

METHODS

Study Design

This was an individual, stratified randomized controlled trial (RCT), with a 2:2:1 ratio for 3 care management interventions (high-touch, high-tech, and ODP, respectively) delivered from September 10, 2018, through November 20, 2022, by the UPMC Health Plan (HPLAN), a large payer within an integrated delivery and finance system in Pennsylvania. Research activities were approved by the University of Pittsburgh’s Institutional Review Board and overseen by a data safety monitoring board. Data were collected via surveys and claims submitted to HPLAN and Pennsylvania Department of Human Services. Trial details, including planned analysis and adaptations, are available in eAppendix 1 (eAppendices available at ajmc.com). Adaptations due to the COVID-19 pandemic include additional sensitivity analyses, sample size reduction, and the allowance for initial visits to be telephonic due to restrictions on in-person visits.

Inclusion and Enrollment

This study enrolled HPLAN Medicaid or dually eligible beneficiaries 21 years and older who were identified by a population health management risk stratification tool as having high and/or rising health care needs, including being at risk for unplanned care use and living with MCC as defined by at least 1 physical health condition and at least 1 additional physical or behavioral health condition, and who were discharged from an inpatient hospitalization within 30 days. Care managers (CMs) called eligible members to set up an initial in-person care visit to complete care assessments, including medication reviews. During this initial visit, eligible participants were approached about study enrollment and randomly assigned after providing informed consent.

Interventions

The UPMC Community Team (CT) has been providing in-person and telephonic integrated care management services to HPLAN members since 2012. The CT is comprised of HPLAN-employed registered nurses and licensed clinical social workers/counselors (CMs). CT services include developing and monitoring individualized chronic disease self-management and self-care support plans, providing patient education, and facilitating linkages to medical, behavioral health, and social services. Further, CMs engage in care coordination with primary and specialty providers. To better understand how to optimize CT delivery and expand its impact, we compared 3 different ways of delivering this service. Our goal was to identify which approach works best for which subpopulations and to explore opportunities for extending the reach of services.

After the initial intake appointment, high-touch and high-tech care management approaches continued for at least 4 months and up to 12 months if needed. High-touch care management was delivered in home or community settings, with more frequent telephonic contacts as needed/requested. High-tech care management was delivered via a remote patient monitoring (RPM) platform, with telephonic contacts as needed. RPM tools included video visits and condition-specific text message check-ins that allow CMs to monitor and address concerns related to care coordination (eg, appointment scheduling), symptom management (eg, abdominal pain, feeling anxious), and/or biometric feedback (eg, glucose level) in real time. High-tech participants received text prompts, ranging from daily to biweekly, tailored to their primary condition(s). Smartphones, data plans, and technical support were provided. ODP included 2 weeks of telephonic care management, with an additional 2 weeks for resource connection and seamless transition. (See eAppendix 2 for intervention details.)

Outcomes

We had 3 primary outcomes: 90-day hospital readmissions, as defined by the all-cause readmission rate for inpatient physical and behavioral health service use, including prescheduled hospitalizations and excluding childbirth, within 90 days of the index admission; health status, as defined by the 36-Item Short Form Health Survey instrument¹⁵; and patient activation, as defined by the Patient Activation Measure.¹⁶

Secondary outcomes included 30-day hospital readmissions; functional status (PROMIS Physical Function Short Form 6b)¹⁷; quality of life (Quality of Life Enjoyment and Satisfaction Questionnaire)¹⁸; care satisfaction (Patient Assessment of Chronic Illness Care survey)¹⁹; emergent care use, as defined by the frequency of emergency department (ED) visits in 12 months; engagement in primary, specialty, and mental health care, as defined by the frequency of visits in 12 months; and gaps in care, as defined by Healthcare Effectiveness Data and Information Set quality measures for asthma, cardiovascular disease, diabetes, and depression, and for congestive heart failure, a 30-day condition-specific hospitalization rate definition was used. Self-reported survey data were collected at baseline and at 3, 6, and 12 months via phone, paper, or email. Claims-based outcomes were extracted at 6-month intervals for a 12-month period.

Sample Size

Primary outcomes were powered at more than 80% to detect small mean and overall differences across the 3 interventions, as estimated by a 2-sample t test for comparing high-touch and high-tech approaches, and, secondarily, using an analysis of variance or χ² test as appropriate for overall differences. We defined 8 subgroups, a priori, for the heterogeneity of treatment effect (HTE) analysis: participants at least 60 years and participants younger than 60 years; Black/African American participants and White participants; participants with a Charlson Comorbidity Index (CCI)²⁰ score of 5 or greater (indicating higher illness complexity) and participants with a CCI score less than 5; and participants with comorbid behavioral and physical health conditions and patients with physical health conditions only. Power estimates ranged from 72.2% to 99.9% and are detailed in the original protocol.

Randomization

We employed an individual, stratified randomized design with a 2:2:1 ratio for high-touch, high-tech, and ODP approaches, respectively, using random block sizes of 5 and 10 to maximize balance between groups. Randomization was stratified by sex, insurance type, and technology comfort. REDCap Cloud 1.5 (REDCap Cloud) autogenerated group assignment based on these parameters.

Statistical Analysis

Following the intent-to-treat principle, we used 2 modeling approaches: a generalized linear mixed model (GLMM) to analyze repeated outcomes and a generalized linear model for single time point outcomes, using logistic, linear, Poisson, or negative binomial (with overinflated variance) regression depending on the outcome distribution. For each analysis, we fit multiple regression models with treatment group as the key independent variable of interest and included main effects age, race, sex, insurance type, illness complexity, socioeconomic status, baseline engagement in the interventions, social support, health literacy, and technology literacy. For longitudinally measured outcomes, the main effect of time and the interaction effect between time and treatment was first evaluated to see whether time effects differed among the interventions (time by treatment); pairwise treatment group effects on changes from baseline at each time point were subsequently assessed if the interaction was significant.

To evaluate HTE across subgroups, we analyzed a 3-way interaction (time-treatment-subgroups) for repeated outcomes and a 2-way interaction (treatment-subgroups) for single time point outcomes. Significance tests from all regression models were conducted initially at P less than .05, but for outcomes with at least 1 significant difference, we used the Benjamini-Hochberg method to control the false discovery rate of 0.05 across the multiple pairwise comparisons.

Primary analyses were conducted under the “missing at random” assumption. We used multiple imputation to explore the impact of missing data on the health status and patient activation; missingness was extremely low for 90-day readmission. We also compared pre– vs post–COVID-19 (ie, collected before vs after March 17, 2020) data to compare demographics, primary outcomes at baseline and over time, and missing data patterns. The statistician was blind to intervention allocation.

RESULTS

Of the 1414 participants randomly assigned, 1387 were eligible for inclusion in the analysis after 24 were found to have been ineligible at the time of randomization and 3 subsequently asked for their data not to be utilized for research purposes (Figure). Of the 1387 participants, the majority were female (63.0%), were Medicaid beneficiaries (79.5%), and had comorbid behavioral and physical health conditions (58.0%); they had a mean CCI of 5.06 and a mean age of 53 years. Participants primarily identified as White (73.0%) or Black/African American (21.6%). Study interventions had similar characteristics across demographic variables (Table 1).

Effectiveness of Interventions

Overall, 85.2% of the 1387 participants had no readmission within 90 days of the index date (ie, last discharge date prior to study enrollment), 14.2% had at least 1 readmission, and 0.6% (n = 8) had missing data. Readmission was tested as a binary outcome because only 31 participants had more than 1 readmission. See eAppendix 3 for empirical proportion of rates. Readmission results did not support our initial hypothesis (H1a). No marginal differences in 90-day readmission rates were observed between the interventions (P = .59), and we found no treatment effect evidence for 90-day readmission rates (P = .669).

Although we hypothesized (H1b) that high-touch would lead to higher health status compared with high-tech and ODP approaches, we found no significant time-by-treatment interaction for health status (effect size, 0.0308; 95% CI, 0.002-0.064; P = .851). See eAppendix 4 for least squares estimates. We did find a positive increasing pattern of scores (significant change over time) for health status (P < .0001), with no significant difference by approach (P = .87). High-touch scores indicated a more favorable health state compared with high-tech and ODP scores; ODP had the lowest health state scores, although associations were not significant. See Table 2 for test of pairwise group difference in the change score and Table 3 for adjusted P values for select outcomes.

The hypothesis that high-tech would result in higher patient activation (H1c) was not confirmed. Although we did observe a significant time-by-treatment-group interaction for patient activation (effect size, 0.0537; 95% CI, 0.007-0.102; P = .02) in the regression model, this result requires caution when interpreting because there was no consistent, monotonic pattern in the estimates of changes. See eAppendix 5 for least squares estimates and effect size. Compared with ODP, the high-touch approach had a statistically significant change in patient activation from baseline to 12 months (P = .03), with an adjusted group difference within the standard minimally clinical difference (2.69 points; SE = 1.22; 95% CI, 0.29-5.09).^21,22 See Table 4 for test of pairwise group difference in the change.

Secondary Outcomes

Of the 1387 participants, 96.3% had no readmission within 30 days of the index date and 3.8% had at least 1 readmission (2 had more than 1 readmission). No marginal differences in 30-day readmission rates were observed between approaches (P = .58). We found no evidence of a treatment effect (P = .744).

There was no significant time-by-treatment interaction for 12-month outcomes for the following: primary care provider (PCP) visits (P = .823), specialist visits (P = .473), inpatient admissions (P = .980), and mental health care visits (P = .962). PCP visits (P < .0001), specialist visits (P < .0001), and mental health care visits (P = .038) significantly increased over time, whereas inpatient admissions (P < .0001) significantly decreased over time. See eAppendix 6 for a test for group differences in change from baseline for PCP visits and specialist visits and eAppendix 7 for OR of treatment and time for inpatient admissions and mental health care visits.

There was a significant time-by-treatment interaction for ED visits (P = .041). High-touch estimated mean ED counts were greater than those for high-tech and OPD approaches at both time points and were particularly greater than the high-tech approach at 12 months (estimated difference, 0.19). High-tech and ODP approaches did not show significant differences at 12 months and 6 months, respectively. See eAppendix 8 for test of pairwise group difference in change score.

There was no significant time-by-treatment interaction for functional status (P = .556), quality of life (P = .520), or care satisfaction (P = .188). Although we did observe a statistically significant increase over time for functional status (P < .0001), quality of life (P < .0001), and care satisfaction (P = .009), there was no significant difference by intervention for any of these 3 outcomes. See eAppendix 9 for test of pairwise group difference in change score.

Approximately 10% of the total sample was eligible for inclusion in the gaps-in-care outcome analyses. There were no statistically significant treatment effects. See eAppendix 10 for the rates of events and the results of the Fisher exact test (eg, without covariate adjustment).

HTE Analysis

For 90-day readmissions and health status, there was no statistically significant 2-way interaction when treatment was interacted with (1) age as a 2-level variable (P = .13; 0.87); (2) race as a 2-level variable (P = .75; 0.81); (3) CCI as a 2-level variable (P = .21; 0.33); and (4) comorbid behavioral health conditions as a 2-level variable (P = .56; 0.47). For patient activation, there was no statistically significant 3-way interaction when time-by-treatment interaction term was further interacted with (1) age as a 2-level variable (P = .45); (2) race as a 2-level variable (P = .76); (3) CCI as a 2-level variable (P = .34); and (4) conditions as a 2-level variable (P = .68).

Sensitivity Analysis

The marginal distribution of the completion of the self-reported survey outcomes was significantly different among the 3 interventions. We analyzed the association between treatment groups and baseline covariates in relation to the completion of 2 primary survey outcomes using logistic regression. After adjusting for the effect of covariates associated with survey completion, we found the outcome missing rates were no longer significantly associated with intervention groups (P = .46). The participants who were engaged at baseline (OR, 1.35), were older (OR by 1-year increments, 1.02), were non-Hispanic (OR, 0.49), and were more comfortable with technology (disagree vs disagree strongly: OR, 0.74; agree vs disagree strongly: OR, 1.00; agree strongly vs disagree strongly: OR, 1.49) and had a lower CCI score (OR, 0.95) tended to have a higher completion adherence for survey measures. Additional analysis that included complete cases only and used 30 times multiple imputations for missing outcomes also showed that our primary inference for the association between intervention and patient activation or health status score changes over time was not sensitive to the missing data after adjusting for the effect of regression covariates. See eAppendix 11 for missingness.

Pre–COVID-19, 31.8% of participants responded to the primary surveys and the proportions were similar among interventions. We no longer found a significant effect for patient activation when only analyzing pre–COVID-19 participants due to a smaller sample. However, when the GLMM for the entire sample was extended to include interactions with COVID-19 indicators and allowing for heteroscedasticity (variance inflation), the inference remained the same as the initial analysis. Health status and 90-day readmission results remained the same.

DISCUSSION

Most published RCTs focus on integrated care management for individuals with specific or singular chronic conditions and do not address the multiplicity and complexity of a patient’s conditions and health care navigation burdens.^11,12 We focused on the effect of 3 care management approaches on several patient-centered outcomes for Medicaid and dually eligible beneficiaries living with MCC, with CMs taking a holistic approach to working with patients to concurrently manage their conditions and social needs.²³

This study examined the impact on several patient-centered outcomes that are typically not collected by payers.²⁴ We found a significant difference between interventions on patient activation, with the high-touch approach having a clinically significant higher positive change from baseline to 12 months compared with ODP.^21,22 This may be due to the nature of the high-touch approach (eg, more time spent with a CM, including more in-person contacts) compared with ODP and high-tech delivery. Health status, functional status, quality of life, and care satisfaction significantly improved over time, with no significant differences across interventions. Further, no significant differences were found for outcomes based on age, race, CCI, and comorbid behavioral health conditions, providing evidence that technology-supported interventions do not risk contributing to care inequities.²³

Like other trials focused on the provision of integrated approaches for chronic conditions,²² we had mixed results on care utilization outcomes. We found no intervention effect for 30- and 90-day readmissions, and although primary, specialty, and mental health care utilization significantly improved over time, there were no differences across interventions. The observed 90-day readmission rate was 14%. This was lower than the expected rate of 20% to 28% that was observed for this population during the design phase. An ad hoc analysis will be conducted to explore utilization rates and costs compared with HPLAN members receiving telephonic-only or no care management services. As health care systems explore care management programs that address MCC, these results suggest that members/patients and CMs can use various modalities, including digital health tools and shorter-duration programs, without compromising outcomes.

Limitations

Several limitations exist. ODP was an active intervention and likely more coordinated than true usual care. Using an active intervention, however, was consistent with our focus on comparative effectiveness and was a mandate of trial design from the sponsor, the Patient-Centered Outcomes Research Institute. This intent-to-treat analysis comprised all participants, including those whom the CT was unable to reach and those who declined services. Additional analysis examining the role of engagement is warranted. Future research should be conducted that considers time variances in illness complexity in relation to intervention effects. Additionally, CMs may have delivered all 3 interventions across their caseloads, which is representative of how care management teams are organized, providing multiple interventions to ensure a patient-centered approach to care. The risk of contamination was mitigated by service guidelines being similar across interventions. The study was underway when the COVID-19 pandemic began, and although we considered the pandemic in our analyses, it is possible that the pandemic influenced results in unmeasurable ways. Moreover, all-cause readmission outcomes included elective admissions due to limitations in accessing scheduling data; the estimated readmissions rate therefore overestimated readmissions strictly from complications. However, our exclusion criteria for people undergoing hemodialysis and active cancer treatment were designed to mitigate this concern. In addition, the randomized study design should avoid any bias in estimating intervention effectiveness. Lastly, not all substance use diagnoses or behavioral health claims may have been included in our analysis due to privacy and data sharing restrictions.

CONCLUSIONS

Health care systems exploring ways to enhance care management services for Medicaid and dually eligible beneficiaries with MCC may use these results to guide equitable implementation of interventions to best suit member/patient preferences, staffing structures, digital resources, and populations served. Given the challenges of managing populations with complex medical and social needs amid shrinking care management resources, results from this study can inform design of effective programs. The interventions studied here incorporated person-centered care tenets, leveraged digital care delivery, and included concentrated yet intensive engagement approaches to enable broad scaling to achieve positive member/patient outcomes.

Acknowledgments

The authors thank all study participants, care managers, UPMC Center for High-Value Health Care quantitative researchers, clinical leadership, the Patient Partners Work Group, the Stakeholder Advisory Board, and the Data Safety and Monitoring Board for their collaboration and input into each stage of this study. The authors also thank the study team for their dedication and preserving adaptability.

Author Affiliations: UPMC Center for High-Value Health Care (KW, JNK, SM), Pittsburgh, PA; Department of Psychiatry, University of Pittsburgh (CK), Pittsburgh, PA; University at Buffalo (DL), Buffalo, NY; UPMC Health Plan (EB, DS), Pittsburgh, PA; now with Partners in Care Foundation (DS), San Fernando, CA; Department of Psychiatry, University of Arizona College of Medicine (JFK), Tucson, AZ.

Source of Funding: This study was partially funded by the Patient-Centered Outcomes Research Institute (PCORI), contract IHS-1609-36670. All statements in this report, including its findings and conclusions, are solely those of the authors and do not necessarily represent the views of PCORI or its board of governors or methodology committee. Financial support not included in the PCORI contract (ie, intervention delivery costs) was provided by the UPMC Health Plan.

Author Disclosures: Dr Williams, Dr Kogan, and Ms Markwardt are employed at a nonprofit within the health system where the study occurred. Dr Beckjord is employed by the health plan that delivers these services. Dr Karp has received consulting or advisory payments from Otsuka and J&J Neuroscience. Dr Swayze was previously employed by the health plan that delivers these services. The remaining authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Authorship Information: Concept and design (KW, JNK, SM, CK, DL, EB, JFK, DS); acquisition of data (KW, SM, DS); analysis and interpretation of data (KW, JNK, SM, CK, DL, DS); drafting of the manuscript (KW, JNK, SM, CK, DL, EB, JFK, DS); critical revision of the manuscript for important intellectual content (JNK, EB, JFK, DS); statistical analysis (CK); provision of patients or study materials (SM, DS); obtaining funding (JNK, CK); administrative, technical, or logistic support (KW, SM, DS); and supervision (KW, JNK, DL, DS).

Address Correspondence to: Kelly Williams, PhD, UPMC Center for High-Value Health Care, US Steel Tower, 600 Grant St, 25th Floor, Pittsburgh, PA 15219. Email: williamsk17@upmc.edu.

REFERENCES

1. Boersma P, Black LI, Ward BW. Prevalence of multiple chronic conditions among US adults, 2018. Prev Chronic Dis. 2020;17:E106. doi:10.5888/pcd17.200130

2. Egede LE. Major depression in individuals with chronic medical disorders: prevalence, correlates and association with health resource utilization, lost productivity and functional disability. Gen Hosp Psychiatry. 2007;29(5):409-416. doi:10.1016/j.genhosppsych.2007.06.002

3. Hewner S, Casucci S, Castner J. The roles of chronic disease complexity, health system integration, and care management in post-discharge healthcare utilization in a low-income population. Res Nurs Health. 2016;39(4):215-228. doi:10.1002/nur.21731

4. CMS Framework for Health Equity 2022-2032. CMS; 2022. Accessed March 25, 2024. https://web.archive.org/web/20250106204212/https://www.cms.gov/files/document/cms-framework-health-equity-2022.pdf

5. Vogeli C, Shields AE, Lee TA, et al. Multiple chronic conditions: prevalence, health consequences, and implications for quality, care management, and costs. J Gen Intern Med. 2007;22(suppl 3):391-395. doi:10.1007/s11606-007-0322-1

6. Savitz LA, Bayliss EA. Emerging models of care for individuals with multiple chronic conditions. Health Serv Res. 2021;56(suppl 1):980-989. doi:10.1111/1475-6773.13774

7. Lockwood C, Mabire C. Hospital discharge planning: evidence, implementation and patient-centered care. JBI Evid Synth. 2020;18(2):272-274. doi:10.11124/JBIES-20-00023

8. Lorig KR, Sobel DS, Stewart AL, et al. Evidence suggesting that a chronic disease self-management program can improve health status while reducing hospitalization: a randomized trial. Med Care. 1999;37(1):5-14. doi:10.1097/00005650-199901000-00003

9. Lorig KR, Ritter P, Stewart AL, et al. Chronic disease self-management program: 2-year health status and health care utilization outcomes. Med Care. 2001;39(11):1217-1223. doi:10.1097/00005650-200111000-00008

10. Wheeler JR, Janz NK, Dodge JA. Can a disease self-management program reduce health care costs? the case of older women with heart disease. Med Care. 2003;41(6):706-715. doi:10.1097/01.MLR.0000065128.72148.D7

11. Bodenheimer T, Wagner EH, Grumbach K. Improving primary care for patients with chronic illness: the chronic care model, part 2. JAMA. 2002;288(15):1909-1914. doi:10.1001/jama.288.15.1909

12. Fortin M, Stewart M. Implementing patient-centered integrated care for multiple chronic conditions: evidence-informed framework. Can Fam Physician. 2021;67(4):235-238. doi:10.46747/cfp.6704235

13. Global Data Plc. The Complexities of Physician Supply and Demand: Projections From 2021 to 2036. Association of American Medical Colleges; 2024. Accessed March 25, 2024. https://www.aamc.org/media/75236/download

14. Bodenheimer T. Strategies to reduce costs and improve care for high-utilizing Medicaid patients: reflections on pioneering programs. Center for Health Care Strategies. October 2013. Accessed March 25, 2024. https://www.chcs.org/resource/strategies-to-reduce-costs-and-improve-care-for-high-utilizing-medicaid-patients-reflections-on-pioneering-programs/

15. McHorney CA, Ware JE Jr, Lu JF, Sherbourne CD. The MOS 36-Item Short-Form Health Survey (SF-36): III. tests of data quality, scaling assumptions, and reliability across diverse patient groups. Med Care. 1994;32(1):40-66. doi:10.1097/00005650-199401000-00004

16. Hibbard JH, Stockard J, Mahoney ER, Tusler M. Development of the Patient Activation Measure (PAM): conceptualizing and measuring activation in patients and consumers. Health Serv Res. 2004;39(4, pt 1):1005-1026. doi:10.1111/j.1475-6773.2004.00269.x

17. Jensen RE, Potosky AL, Reeve BB, et al. Validation of the PROMIS physical function measures in a diverse US population-based cohort of cancer patients. Qual Life Res. 2015;24(10):2333-2344. doi:10.1007/s11136-015-0992-9

18. Ritsner M, Kurs R, Gibel A, Ratner Y, Endicott J. Validity of an abbreviated Quality of Life Enjoyment and Satisfaction Questionnaire (Q-LES-Q-18) for schizophrenia, schizoaffective, and mood disorder patients. Qual Life Res. 2005;14(7):1693-1703. doi:10.1007/s11136-005-2816-9

19. Glasgow RE, Wagner EH, Schaefer J, Mahoney LD, Reid RJ, Greene SM. Development and validation of the Patient Assessment of Chronic Illness Care (PACIC). Med Care. 2005;43(5):436-444. doi:10.1097/01.mlr.0000160375.47920.8c

20. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373-383. doi:10.1016/0021-9681(87)90171-8

21. Anderson JK, Wallace LM. Evaluation of uptake and effect on patient-reported outcomes of a clinician and patient co-led chronic musculoskeletal pain self-management programme provided by the UK National Health Service. Br J Pain. 2018;12(2):104-112. doi:10.1177/2049463717734015

22. Schuster J, Reynolds C III, Carney T, et al. Using Wellness Coaches and Extra Support to Improve the Health and Wellness of Adults With Serious Mental Illness. Patient-Centered Outcomes Research Institute; 2019. Accessed March 30, 2024. https://www.ncbi.nlm.nih.gov/books/NBK603250/

23. Sherman BW. Management of individuals with multiple chronic conditions: a continuing challenge. Am J Manag Care. 2021;27(6):256-260. doi:10.37765/ajmc.2021.88665

24. Finkelstein A, Zhou A, Taubman S, Doyle J. Health care hotspotting — a randomized, controlled trial. N Engl J Med. 2020;382(2):152-162. doi:10.1056/NEJMsa1906848