The effect of nurse care management on multimorbid patients’ hospital utilization: a comparative-controlled trial and an exploratory cohort study Authors: Efrat Shadmi, Orly Tonkikh, Marcelo Low, Calanit Key, Doron Netzer, Tchiya Molcho, Shmuel Giveon, Chad Boult, Ran D Balicer Published: 2026 ABSTRACT AbstractMultimorbidity, defined as the presence of two or more chronic conditions in one individual, poses substantial challenges to patients and healthcare systems, yet evidence regarding effective primary care interventions remains limited. We report on a two-year comparative controlled trial aimed to assess the effects of the nurse-led Comprehensive Care for Multimorbid Adults Program (CC-MAP), on patient outcomes. A total of 1,210 adults with multimorbidity were enrolled (610 in seven intervention clinics and 600 in six usual-care clinics). The primary outcome was the number of unplanned admissions. Secondary outcomes included the number of 30-day readmissions, self-reported physical and mental health, and functional status. Adjusted mean differences between groups were estimated using analysis-of-covariance. At one year, no significant between-group differences were observed. At two years, the intervention group had lower adjusted mean unplanned hospital admissions (− 0.468; 95% CI − 0.815 to −0.121) and lower 30-day readmissions (− 0.202; 95% CI − 0.370 to −0.034), relative to controls. Our findings support the potential sustainability and scalability of the CC-MAP primary care model for patients with multimorbidity. (ClinicalTrials.gov (Identifier: NCT01811173)).Subject terms: Health policy, Health services Introduction More than two decades after the systemic burden of multimorbidity, commonly defined as two or more chronic conditions within an individual, was prominently recognized1–6, it continues to pose substantial challenges for patients, families, providers, and healthcare systems worldwide7–9. In response to the challenges of multimorbidity, a range of care models have been tested, with self-management support and care coordination standing out as successful core strategies, particularly emphasizing nurse-led interventions in primary care settings10,11. Evidence suggests that such models can lead to improved mental health and functional outcomes for patients12. Achieving sustainability with multimorbidity programs has remained difficult, however, and the complexity of multimorbidity continues to strain care delivery systems today. Major obstacles have included the challenge of offsetting the costs associated with additional program staffing and determining effective nurse-to-patient ratios to ensure long-term program sustainability11,13,14. One comprehensive care management model that deploys a registered nurse within a primary care team to provide patient monitoring, coaching, self-management support, and care coordination is “Guided Care”15. Informed by the Chronic Care Model16, Guided Care improves patient engagement and adherence to treatment recommendations, and it has been shown to improve the quality of care for chronic conditions17. Guided Care’s favorable effects on hospital days (21% reduction) and skilled nursing facility days (52% reduction) were observed, however, only in American patients who received care from a fixed-budget, payer-provider organization—Kaiser Permanente15,18,19. Building on evidence from studies of Guided Care and related programs, Clalit Health Services (Clalit) designed the Comprehensive Care for Multimorbid Adults Program (CC-MAP) in 2012 and has begun to test its effects and its long-term outcomes. Clalit is Israel’s largest not-for-profit healthcare payer-provider organization, serving over 5 million individuals. The CC-MAP program is a primary care-based model that is led by nurses who work closely with primary care physicians to manage the comprehensive care of high-risk patients with multimorbidity. Clinics and patients are selected for the CC-MAP program through a two-step process. First, clinics are selected for CC-MAP participation if they serve a population with a high morbidity burden, as assessed using the well-validated “Johns Hopkins Adjusted Clinical Groups (ACG®) case-mix system” morbidity scores, which are derived from diagnostic data from all medical encounters20,21. Second, within the eligible clinics, patients are selected if they are at high risk for high healthcare utilization and costs, as defined by an ACG® risk score above a threshold (0.19), which corresponds to the top 6% of each clinic’s patient population in terms of multimorbidity and risk of high future health care utilization. Evidence indicates that individuals at this level of risk and anticipated utilization benefit from proactive care management to improve outcomes and better coordinate care22. Since enrollment of its first patients in 2013, the CC-MAP has demonstrated its feasibility, high uptake, and effectiveness in patient-centered medication management23 and achieving high adherence to care recommendations24. In the present comparative-controlled trial, we evaluated whether the CC-MAP is associated with patients’ health, function and hospital utilization outcomes over time. Specifically, we used an intention-to-treat strategy to assess whether multimorbid patients who enrolled in the CC-MAP, compared to those who received usual care, had fewer unplanned hospital admissions, total inpatient days, and 30-day readmissions during 1 and 2 years following their enrollment. Additionally, we conducted telephonic surveys at baseline and one year later to evaluate the effect of CC-MAP on patients’ self-reported physical and mental health and functioning. To examine the potential scalability of CC-MAP, we then performed a 2-year exploratory cohort study to assess whether CC-MAP patients’ hospital utilization differed when CC-MAP nurses’ caseloads were increased from 100 to 150 patients per nurse. Results Clalit primary care clinics in two districts (n = 30) were eligible for the trial if they served populations with low-medium socioeconomic status and a high morbidity burden according to the ACG® case-mix system20,21. Of these 30, 13 clinics were randomly selected (by random numbering of the 30 clinics and selection of the top 13). Seven clinics were then designated for intervention, and seven CC-MAP nurses were assigned to them (one full-time nurse to each of four clinics and one 75% nurse to each of the other three). Each CC-MAP nurse managed a caseload of 100 patients per full-time equivalent nurse position. The remaining six clinics served as controls. Assignment to the intervention or control arm of the trial was not random, but rather was determined pragmatically, based on the geographic availability of CC-MAP nurses to be integrated into the seven intervention clinics. Although not randomly selected, the intervention and control clinics were comparable in size, geographic location, and the socioeconomic characteristics of the populations they served. Baseline differences between the intervention and control groups were entered as covariates in multivariate models of all outcomes. Six hundred and ten multimorbid patients of the seven intervention clinics and 600 multimorbid patients of the six “usual care” control clinics were eligible, gave consent, and were enrolled in the trial. Eligibility required: multimorbidity (having 2 or more chronic conditions), age 45–95 years, language fluency, and having a high risk for high utilization of healthcare services in the future (as defined by having a baseline ACG® score of >0.19)22. The enrollment and progress of patients through the trial is shown in Fig. 1. The study was determined as completed after the end of the follow-up period for the last recruited patient (June 2017). The CC-MAP CONSORT flow diagram. Table 1 compares the sociodemographic and clinical characteristics of the intervention and control cohorts at baseline. The groups were similar in sex, living arrangements, socioeconomic status, prevalence of several chronic conditions, basic and instrumental activities of daily living, and use of hospital services during the year before enrollment. The intervention group patients were statistically significantly younger than the control patients, less educated, less likely to be recent immigrants, and had lower prevalence of several chronic conditions, including depression, ischemic heart disease, arrythmia, hypertension and chronic renal failure. The intervention group also had lower (worse) mental health scores and slightly higher (better) physical health scores at baseline. Mortality monitoring revealed no significant differences in mortality rates at the 15 and 27-month follow-up periods between groups (for more details, please see the Methods section). There were no significant differences between the patients who did and did not provide complete survey data (Supplementary Table 1). Study participants hold similar characteristics to Clalit members with multiple chronic conditions who are at high risk for deterioration22. Table 1Baseline characteristics and bivariate analysis (Chi-square for categorical variables and two-tailed t-test for continuous variables) of the CC-MAP intervention and control patientsIntervention (n = 610)Control (n = 600)p-valueCategorical variablesn%n%Sex (Female)35658.432954.80.233Education level: Elementary school only33454.828747.80.018Living arrangement: Alone23939.222938.20.724Immigrated since 1990548.97512.50.041Socioeconomic status: Low24039.323439.00.906Chronic conditionsa Malignancy14423.612420.70.239 Diabetes30049.227145.20.167 Obesity24339.825843.00.268 Hyperlipidemia54689.555091.70.203 Depression9916.213021.70.019 Ischemic heart disease23738.927846.30.009 Congestive heart failure508.2569.30.542 Arrhythmia13622.317929.80.003 Hypertension46275.749682.70.004 Chronic obstructive pulmonary disease6911.38313.80.194 Chronic renal failure12520.515225.30.047Hospital utilization in the year before enrollment: One or more 30-day readmissions345.6264.30.355Continuous variablesMeanSDMeanSDAge70.6710.2973.449.9<0.001Number of chronic conditions5.232.535.552.550.029Hospital utilization in the year before enrollment Number of unplanned admissions0.350.700.330.730.479 Total number of inpatient days2.026.131.695.040.305Health and functionb SF-12 physical health34.6510.7232.5911.420.012 SF-12 mental health45.3411.7147.5312.600.016 Activities of daily living5.570.755.570.750.985 Instrumental activities of daily living2.961.903.221.810.052CC-MAP comprehensive care for multimorbid adults program, SD standard deviation, SF-12 12-item Short-form Health survey.a608 patients in the intervention group had complete diagnoses data.b355 patients in the intervention group and 457 in the control group. Baseline characteristics and bivariate analysis (Chi-square for categorical variables and two-tailed t-test for continuous variables) of the CC-MAP intervention and control patients CC-MAP comprehensive care for multimorbid adults program, SD standard deviation, SF-12 12-item Short-form Health survey. a608 patients in the intervention group had complete diagnoses data. b355 patients in the intervention group and 457 in the control group. Hospital utilization Table 2 compares the hospital utilization by the intervention and control groups during the first year and during both years of the trial. One-year outcomes were evaluated over the period spanning 3–15 months following enrollment, while 2-year outcomes encompassed the period from 3 to 27 months post-enrollment. These timeframes were selected to exclude the initial three-month period after enrollment, during which hospitalization outcomes are unlikely to be influenced, given that care plans are still being formulated and patients have not yet begun implementing changes to their treatment regimens. Table 2Bivariate comparison (two-tailed t-test) of hospital utilization during the first (3–15 months) and second (3–27 months) years of follow-up after enrollment by the intervention versus control groups (crude means and standard deviations)Intervention (n = 608)Control (n = 600)p-valueMeanSDMeanSD1-year outcomes Number of unplanned admissions0.330.760.400.880.167 Number of 30-day readmissions0.070.340.090.350.250 Total number of inpatient days1.615.022.537.540.0132-years outcomes Number of unplanned admissions0.691.250.891.450.011 Number of 30-day readmissions0.160.580.220.660.081 Total number of inpatient days3.859.795.6511.550.004SD standard deviation. Bivariate comparison (two-tailed t-test) of hospital utilization during the first (3–15 months) and second (3–27 months) years of follow-up after enrollment by the intervention versus control groups (crude means and standard deviations) SD standard deviation. During the first year of follow-up, there was no statistically significant difference between the control group and the intervention group in the primary outcome of total unplanned admissions. Difference between the groups was observed, with the control group having significantly higher mean total inpatient days than the intervention group (2.53 vs. 1.61 days, p = 0.013). During the whole 2-year follow-up period, the control group had significantly more unplanned admissions (mean = 0.89 vs. 0.69 admissions, p = 0.011) and more total inpatient days (mean = 5.65 vs. 3.85 days, p = 0.004). Using ANCOVA, we then performed adjusted analyses. The covariates in the adjusted models were those in which baseline differences were found between the two groups (see Table 1): age, education level, immigration status, number of chronic conditions, specific chronic conditions (depression, ischemic heart disease, arrhythmia, hypertension, chronic renal failure), and clinic. During the 1-year follow-up period (see Fig. 2), there were no statistically significant differences in the primary outcome of adjusted mean numbers of unplanned admissions (Intervention = 0.30 (95% CI 0.18–0.42), Control = 0.44 (95% CI 0.32–0.56); F(1,1186) = 1.533, p = 0.216), and in total inpatient days (Intervention = 1.44 (95% CI 0.51–2.37), Control = 2.71 (95% CI 1.77–3.65); F(1,1186) = 2.067, p = 0.151) and 30-day readmissions (Intervention = 0.04 (95% CI −0.01–0.09, Contro = 0.12 (95% CI 0.07–0.17); F(1,1186) = 2.917, p = 0.088). The adjusted marginal mean differences between the intervention and control groups in unplanned admissions (−0.139, 95% CI −0.360–0.081), total inpatient days (−1.268, 95% CI −2.998– 0.462), and 30-day readmissions (−0.082, 95% CI −0.176–0.012) were not statistically significant. First year hospital utilization outcomes. Comparison (analysis of covariance) of the intervention (n = 608) and control (n = 600) groups’ adjusted marginal mean number of unplanned admissions (a), 30-day readmissions (b) and total inpatient days (c) within 3–15 months follow-up period, adjusted for age, immigration status, low education, number of chronic conditions, ischemic heart disease, depression, arrhythmia, chronic renal failure, hypertension and clinic. Error bars represent 95% confidence intervals. n denotes the number of participants contributing to the estimates in each group. Source data are provided as a Source data file. As shown in Fig. 3, during the 2-year follow-up period, the control group had significantly higher adjusted mean numbers of unplanned admissions (Intervention = 0.56 (95% CI 0.37–0.75), Control = 1.03 (95% CI 0.84–1.22); F(1,1186) = 7.018, p = 0.008), total inpatient days (Intervention = 2.91 (95% CI 1.41–4.41), Control = 6.61 (95% CI 5.09–8.12); F(1,1186) = 6.710, p = 0.010) and 30-day readmissions (Intervention = 0.09 (95% CI −0.002– 0.18), Control = 0.29 (95% CI 0.20–0.38); F(1,1186) = 5.557, p = 0.019). The adjusted marginal mean differences between the intervention and control groups in unplanned admissions (−0.468, 95% CI −0.815 to −0.121), total inpatient days (−3.695, 95% CI −6.494 to −0.896) and 30-day readmissions (−0.202, 95% CI −0.370 to −0.034) were statistically significant. Second year hospital utilization outcomes. Comparison (analysis of covariance) of the intervention (n = 608) and control (n = 600) groups’ adjusted marginal mean number of unplanned admissions (a), 30-day readmissions (b) and total inpatient days (c) within the 2-year follow-up period, adjusted for age, immigration status, low education, number of chronic conditions, ischemic heart disease, depression, arrhythmia, chronic renal failure, hypertension and clinic. Error bars represent 95% confidence intervals. n denotes the number of participants contributing to the estimates in each group. Source data are provided as a Source data file. Patient-reported outcomes Table 3 shows the secondary outcomes of patient-reported outcomes for the intervention and control patients who responded to the survey 1 year after enrollment. These bivariate comparisons of the intervention and control groups’ responses show significantly better physical health in the intervention group. Table 3Patient-reported outcomes for the intervention and control patients who responded to the survey 1 year after enrollment (crude means, standard deviations and bivariate comparisons using two-tailed t-test)InterventionControlp-valuenMeanSDnMeanSD SF-12 physical health31734.2511.9643631.6111.520.002 SF-12 mental health31746.0913.4043645.4912.640.528 Activities of daily living2865.471.004045.471.020.916 Instrumental activities of daily living3163.001.924272.791.930.145SD standard deviation, SF-12 12-item short-form health survey. Patient-reported outcomes for the intervention and control patients who responded to the survey 1 year after enrollment (crude means, standard deviations and bivariate comparisons using two-tailed t-test) SD standard deviation, SF-12 12-item short-form health survey. Multivariate adjustment of these means showed that there were no significant 1-year differences between the study groups in physical health (F(1,722) = 0.978, p = 0.323), mental health (F(1,722) = 0.379, p = 0.538), activities of daily living (ADL) (F(1,682) = 1.474, p = 0.225) or instrumental activities of daily living (IADL) (F(1,756) = 1.968, p = 0.161) after controlling for the baseline differences between the groups’ scores (Short Form-12 (SF-12) Physical health, SF-12 Mental Health, ADL, and IADL) for each of these four scales, age, education level, immigration status, number of chronic conditions, and specific chronic conditions (depression, ischemic heart disease, arrhythmia, hypertension, chronic renal failure) and clinic. Exploratory cohort study In the exploratory cohort study, we sought to determine whether CC-MAP patients would have similar hospital utilization when the CC-MAP nurse-to-patient ratio was increased from 1:100 to 1:150. To conduct this study, we enrolled a second cohort of 583 multimorbid patients from five additional CC-MAP clinics whose patients had similar sociodemographic and morbidity characteristics and in which the nurse-to-patient ratio was 1:150. The baseline characteristics of this new cohort were similar to those of the intervention group in the comparative-controlled trial, except for sex, socioeconomic status, obesity, Chronic Obstructive Pulmonary Disease (COPD), Chronic Renal Failure (CRF), and unplanned admissions during the year prior to enrollment (see Supplementary Table 2). Table 4 presents bivariate comparisons of the two cohorts’ mean numbers of 1- and 2-year unplanned admissions and total inpatient days. There were no statistically significant differences between the hospital utilization rates of the cohorts with nurse-to-patient ratios of 1:100 and 1:150. Table 4Bivariate comparison of hospital utilization by CC-MAP patients in the controlled trial (n = 608) and the exploratory cohort study (n = 583) during the first (3–15 months) and second (3–27 months) years of follow-up after enrollment (crude means and standard deviations and bivariate comparisons using two-tailed t test)Controlled trial (1:100 nurse/patient ratio)Exploratory cohort study (1:150 nurse/patient ratio)p-valueMeanSDMeanSD1-year outcomesMean number of unplanned admissions0.330.760.390.870.221Mean total number of inpatient days1.615.022.076.330.1702-year outcomesMean number of unplanned admissions0.691.250.831.450.230Mean total number of inpatient days3.859.794.8512.950.199SD standard deviation. Bivariate comparison of hospital utilization by CC-MAP patients in the controlled trial (n = 608) and the exploratory cohort study (n = 583) during the first (3–15 months) and second (3–27 months) years of follow-up after enrollment (crude means and standard deviations and bivariate comparisons using two-tailed t test) SD standard deviation. After adjusting for the two cohorts’ baseline differences in sex, obesity, COPD, CRF and clinic, the mean total inpatient days during the first year of follow-up of patients in the exploratory cohort study, with a nurse to patient ration of 1:150, was significantly greater than that of intervention patients in the controlled trial, with a nurse to patient ration of 1:100 (F(1,1177) = 3.904, p = 0.048). However, the 2-year follow-up data showed no difference between the two groups’ mean total inpatient days (F(1,1177) = 1.690, p = 0.194). The number of unplanned admissions was not statistically significantly different between the two cohorts after 1 year of follow-up (F(1,1177) = 3.278, p = 0.071) or after 2 years of follow-up (F(1,1177) = 3.256, p = 0.071). Discussion To our knowledge, this is one of the largest comparative-controlled trials of a primary care-based multimorbidity intervention’s 2-year impact on hospital utilization. Compared to usual care, the CC-MAP intervention, which emphasizes care coordination and support for patient self-management, was associated with significant reductions in unplanned hospitalizations, total inpatient days and 30-day readmissions during the 2 years after its implementation. Our exploratory cohort study provides additional insight into whether hospital utilization rates can be maintained as CC-MAP nurses’ caseloads increase. Comparable outcomes were achieved despite a 50% increase in nurse caseloads. Taken together, the CC-MAP model’s reduction in hospital utilization and the persistence of these benefits with a 1:150 nurse-to-patient ratio support the potential long-term sustainability of the CC-MAP intervention. Many previous studies have failed to provide robust evidence supporting the effectiveness of multimorbidity care management interventions. The most recent Cochrane review14 of the effectiveness of multimorbidity care management interventions included studies with durations ranging from 325,26 to 18 months15,19,27. Overall, the review concluded that these interventions likely have little or no effect on patient-reported physical or mental health14. Our 1-year findings are consistent with these conclusions: we observed no significant differences in self-reported physical or mental health or basic or instrumental functioning. Our controlled trial’s 1-year, bivariate hospital utilization analysis shows an annual average difference of almost one inpatient day per person between intervention and control patients. This difference, however, did not remain statistically significant after controlling for all covariates that were unequal at baseline. In contrast, the 2-year adjusted analysis shows significant differences between the intervention and control groups in the mean number of unplanned hospitalizations, the mean total number of inpatient days and in the mean number of 30-day readmissions. The absence of positive outcomes at 1 year, contrasted with the significant hospital utilization benefits we observed after 2 years, underscores the importance of allowing sufficient time for complex care interventions to influence utilization outcomes in patients with multimorbidity. This time-sensitivity of achieving measurable benefits is consistent with findings from other complex care and patient-engagement studies, where establishing trust and effective working relationships is considered foundational and often develops gradually over time, particularly in multimorbidity care models27. A recent overview of multimorbidity care interventions reinforces this perspective, highlighting that the typical 1-year follow-up period used in most studies, largely due to practical and financial constraints, precludes the ability to assess the longer-term impacts of such interventions, particularly those that seek to improve slowly incremental processes like patients’ health behaviors, the delivery of primary care, and the management of patients’ medications28. In line with this rationale, we began measuring patients’ hospital utilization outcomes three months after their enrollment in our studies, considering that neither behavioral changes by patients nor care changes by the healthcare system occurs immediately. Similar future studies should consider incorporating such a “measurement lag” to account for the complexity of the intervention and the gradual, incremental nature of its intended impact. Our exploratory cohort study also contributes to current knowledge by assessing the CC-MAP’s ability to suppress hospital utilization despite a significant increase in the average nurse caseload. Our finding of no difference in hospital utilization outcomes after an increase in caseload, indicates that a 1:150 CC-MAP nurse-to-patient ratio may be feasible for limiting hospital utilization. Similar care management programs typically maintain substantially lower nurse caseloads; for example, the Guided Care model assigned approximately 60–70 patients per nurse29. However, optimal caseload thresholds are highly context-dependent and influenced by the specific roles and responsibilities assigned to the care managers30. In the CC-MAP program, nurses did not conduct home visits, and the patients were usually fully independent in managing their daily activities. These factors have likely contributed to the scalability of the CC-MAP. Nonetheless, the fact that, at 1 year, the exploratory cohort study group (1:150 nurse-to-patient ratio) had a slightly higher adjusted average total number of inpatient days than that of the intervention group in the controlled trial (1:100 nurse-to-patient ratio) may suggest that with a greater caseload, nurses may face challenges in initiating care for all new patients and achieving early stabilization. Indeed, anecdotal reports from our CC-MAP nurses indicate that their most demanding period is during patient enrollment. As outlined above, the CC-MAP model incorporated several strategies to enhance its feasibility and scalability. A critical factor for the sustainability of multimorbidity care management programs is having an effective method for identifying at-risk patients who are most likely to benefit from this extra care31,32. We used the ACG® risk calculator to analyze data from the Clalit database to identify clinics that were highly burdened with multimorbidity and to allocate to them nurses who were trained and dedicated to managing high-risk multimorbid patients in collaboration with their primary care physicians. Additionally, the embedding of our program within patients’ usual primary care clinics, combined with active coordination between the CC-MAP nurse and each patient’s regular primary care physician, likely contributed to the relatively high uptake and sustained engagement observed in the program, as we have previously reported24. This positive feature should be considered when implementing future multimorbidity care management programs, as embedding care models within patients’ usual primary care contexts and aligning team-based care and coordination with existing workflows has been recognized as a core element of effective multimorbidity care delivery and supports uptake and sustainability33. Furthermore, by focusing on clinics that primarily served low socioeconomic populations, the CC-MAP intervention was inherently directed toward individuals with complex social as well as medical needs. CC-MAP nurses identified patients’ social and welfare needs through structured assessments and provided referrals to community services, such as meal services and older adults’ day centers. For patients identified as entitled to social and financial benefits, social services were contacted and provisions were made to ensure they received needed social support. This focus has potentially contributed to the success of the CC-MAP as social deprivation is a well-established contributor to multimorbidity and to the challenges of its management28,34. Prior research has demonstrated the positive impact and cost-effectiveness of other care interventions that have been implemented in socioeconomically disadvantaged settings35. More than two-thirds of these studies’ patients were living alone, a known risk factor for unplanned hospitalizations among individuals with multimorbidity36. Through our multi-level targeting strategy, we were likely able to identify clinics and patients who were most likely to benefit from the intervention37, which may have contributed to its observed effectiveness. Our study has several limitations. First, CC-MAP has been a 14-year program, and despite Clalit’s consistent primary care orientation, some aspects of our comparative-controlled trial data may not reflect all of these clinics’ current clinical practices. Nonetheless, our subsequent exploratory cohort study and its 27 months of follow-up reflect the long-term nature required for evaluating complex care interventions for patients with multimorbidity. Importantly, early positive findings from the comparative-controlled trial informed Clalit Health Services’ decision to scale up the intervention; currently, approximately 3000 patients are enrolled in CC-MAP across more than 20 primary care clinics in Israel. This wide-scale adoption, alongside our results, supports the program’s feasibility and sustainability. Second, we were unable to randomize at the patient level for the comparative-controlled trial, as each clinic provided similar care (either “usual care” or CC-MAP care) for all its multimorbid patients. Consequently, some baseline differences were observed between the trial’s intervention and control groups, as well as between the trial’s and the cohort study’s CC-MAP patients. We addressed these imbalances as completely as possible through multivariate adjustments in our statistical analyses. Third, the original protocol specified admissions for ambulatory care sensitive conditions (ACSCs) as a primary outcome, yet, due to the absence of data on patients’ primary reasons/diagnoses for hospital admissions, we could not identify ACSC-related hospitalizations. Consequently, we used overall unplanned hospital admissions as the primary utilization outcome. Although this substitution may have reduced outcome specificity, unplanned admissions may represent a more appropriate outcome, given recent evidence questioning the validity of ACSCs as reliable indicators of avoidable hospitalizations38. Fourth, survey response rates at baseline and follow-up were moderate (67%), which may have limited our power to detect differences in patient-reported outcomes. Fortunately, we found no evidence of systematic bias between the respondents and the non-respondents, suggesting that the response rates had minimal impact on our findings. Future research on care management for multimorbid patients should evaluate patient-reported health and functioning over two or more years to determine whether changes in these crucial variables, like our observed 2-year reductions in unplanned hospitalizations, total inpatient days and 30-day readmissions, require sustained interventions and follow-up longer than 1 year. This study provides important insights into the long-term management of patients with multimorbidity in primary care. Beyond demonstrating that the intervention was associated with significantly lower rates of unplanned hospital admissions, total inpatient days and 30-day readmissions, our findings suggest the possibility of achieving these outcomes with higher nurse caseloads. Critically, the results underscore the necessity of extended follow-up periods to fully capture the impact of complex care interventions, an important lesson for future studies internationally. Furthermore, by targeting socially and medically vulnerable populations, the CC-MAP model illustrates the value of addressing social determinants of health as part of comprehensive care management. Taken together, these lessons offer valuable lessons for designing sustainable, scalable multimorbidity interventions across diverse healthcare settings. Methods We conducted a comparative-controlled trial and a subsequent exploratory cohort study to evaluate the effects of the CC-MAP intervention. The study was approved by the Clalit Research Ethics Committee (03/2013) which is the authorizing body for all locations of primary care clinics included in the study, and by the Ethics Committee of the Faculty of Social Welfare and Health Sciences at the University of Haifa. There was no direct patient or public involvement in the design or reporting of the study. For the comparative-controlled trial, 1210 multimorbid patients from 13 primary care clinics were enrolled between April 3, 2013, and March 19, 2015, and were followed until June 2017. The 13 clinics and their multimorbid patients were allocated to either the “CC-MAP” intervention cohort (7 clinics) or a “usual care” control cohort (6 clinics) by the research team. Assignment to the intervention and control arms was not randomized or masked. Rather, pragmatic allocation of clinics to intervention and control arms was made, based on limited nurse availability to work in CC-MAP clinics. Intervention and control clinics were comparable in size, location, and socioeconomic characteristics of the populations served. In each intervention clinic, a specially trained nurse collaborated with 3–4 primary care physicians to provide comprehensive care for about 100 high-risk multimorbid patients. In each control clinic, 3–4 primary care physicians provided usual care for about 100 high-risk multimorbid patients. This trial is registered at https://clinicaltrials.gov/study/NCT01811173?intr=NCT01811173&rank=1, (trial number NCT01811173; Registration Date: February 2013). The original protocol included three arms: the intervention and control arms reported above, and a third covert control arm in which patients were not actively recruited, and only prospective follow-up of their data was planned. Due to regulatory changes, in line with the requirements of the Institutional Review Board, for the third covert control arm in which patients were not actively recruited—data was not collected or analyzed. For the subsequent exploratory cohort study, 583 multimorbid patients from five additional clinics were enrolled between March 13, 2016, and February 12, 2018, and received CC-MAP care through May 2020. This study was conducted following the adoption of CC-MAP as the standard of care in primary care clinics across three Clalit Health Services districts, including all the primary care clinics that participated in this exploratory cohort study. In these CC-MAP clinics, each nurse collaborated with 3-4 primary care physicians in managing 150 (rather than 100) multimorbid patients. Setting The studies involved large primary care clinics (5000–10,000 patients/clinic) that served socio-economically disadvantaged populations. Israel’s National Health Insurance Law39 mandates universal health coverage for all residents and requires them to be registered with a primary care physician within one of Israel’s four national health funds’ primary care clinics. Clinics in Clalit Health Services, the largest health fund, were selected for inclusion in the study if they served a population with a high multimorbidity burden. Such burden was designated if the overall morbidity burden weighted sum of the patient population of at least 4 primary care physicians in the clinic exceeded the total count of patients in their panels by 200 (i.e., for primary care physicians with a patient panels of 2000 adult patients, the average weighted morbidity score exceeded 2200, indicating that, on average, their patients were sicker than the average Clalit patient). Morbidity burden was measured using the Johns Hopkins Adjusted Clinical Groups System (ACG®) case mix system, which uses diagnostic codes to stratify individuals according to their morbidity. These groupings reflect cumulative disease burden, clinical complexity, and anticipated healthcare resource use21. Primary care physicians serve as gatekeepers to most services, and all healthcare information from all medical encounters is documented in electronic health records and stored in a centralized data warehouse40. Intervention The CC-MAP follows the principles of the Guided Care model29 a proactive, nurse-led, primary care, team-based model of care designed to improve outcomes for chronically ill, multimorbid older adults. Guided Care integrates principles of the Chronic Care Model and emphasizes coordination across care settings. Like Guided Care, the CC-MAP comprises 8 component activities: (1) Comprehensive patient assessments, (2) Generation of proactive care plans (for providers) and consistent “action plans” (for patients), (3) Support for chronic disease self-management by patients, (4) Monthly-to-quarterly patient monitoring sessions, (5) Patient coaching via motivational interviewing techniques, (6) Coordination of care across settings, especially in cases of hospitalization, (7) Support for family caregivers, and (8) Facilitating patients’ access to community resources. CC-MAP nurses were recruited from the Clalit community care nursing staff and received targeted training in the CC-MAP model. Special training emphasis was placed interdisciplinary teamwork, caring for adults younger than 65 years old, and acquiring expertise with healthcare resources and tools, such as using the electronic health record for care management. Seven nurses (6 FTE) were integrated into the seven intervention clinics. At each intervention clinic, the CC-MAP primary care physicians participated in two full-day CC-MAP training sessions, and the program was introduced to all staff members (additional physicians, nurses, and administrative staff) at a dedicated kick-off meeting and at specific clinic meetings. Since the intervention was customized to each patient’s needs and preferences, the intensity and frequency of services and contacts varied across individual patients. The nurses worked closely with the primary care physicians to generate the patients’ integrated care plans, and they identified and connected patients with additional health and social care services and resources as needed. During the first year, patients met with the CC-MAP nurse in the primary care clinic for face-to-face visits at least once per month, supplemented by telephone contacts or additional clinic visits as needed. After the first year, visit frequency was adjusted according to the individualized care plan and the patient’s healthcare needs, with a minimum of one visit per quarter. Shortly after each patient’s enrollment, the nurse met with the patient’s primary care physician to establish a care plan; subsequently they met periodically (quarterly or less), depending on the patient’s clinical status and the need for care plan adjustments. Although enrollment was voluntary, CC-MAP nursing services were complementary to patients’ routine primary care and, once initiated, participation generally continued throughout the follow-up period. The intensity of services was adjusted according to patients’ needs and clinical status, and participation was discontinued only in the event of death. In a previous study, we showed that CC-MAP nurses applied a defined set of key care processes based on abbreviated guideline references for common chronic conditions, integrating condition-specific recommendations into a comprehensive, patient-centered care plan24. Overall concordance with guideline recommendations was high, with nearly 80% of care delivered in accordance with guidelines across the study population. Deviations from guideline recommendations primarily reflect clinically appropriate adjustments made by providers, such as to address medication interactions or conflicting recommendations. Importantly, patient adherence to referred care exceeded 90%, indicating high uptake and acceptability of the program. Recruitment and data collection procedure Patients were drawn from the primary care populations of the selected clinics. Inclusion criteria required patients to be aged 18 and older, fluent in Hebrew or Russian, diagnosed with two or more chronic conditions (as specified below under Baseline Characteristics), and at high risk for high utilization of healthcare. The ACG® high-utilization risk score of 4% highest risk (ACG® score > 0.19), was set as the threshold, fixed at baseline, based on a prior validation study showing that this threshold identifies high-risk, multimorbid patients who might benefit from a multimorbid care management program22. Exclusion criteria included: patients with major active mental illness, such as schizophrenia, patients with major cognitive failure, patients undergoing dialysis or oncology patients receiving chemotherapy, radiotherapy or other oncology treatment during the past 3 years, kidney, liver, or heart transplant recipients, patients already included in a controlled disease management program (e.g., COPD disease management and telehealth), and non-community-dwelling individuals (inpatient nursing care, nursing homes), participation in any medical research, patient who were housebound or confined to bed, Clalit Healthcare Services employees, and patients who were non Hebrew speakers and without Hebrew speaking primary informal caregiver. Inclusion and exclusion criteria were similar in the exploratory cohort study. All participants in the trial provided written informed consent. No compensation was provided for participants. The patient-reported outcome data were collected from the patients in the intervention and control cohorts upon enrollment and 1 year later by an external survey service via telephone. Data on demographics, chronic morbidity and hospital utilization were retrieved from the Clalit data warehouse for participants in both studies. Harms were defined as serious adverse events (SAE) and included only mortality (as hospitalizations, which can also be considered SAE, were the primary outcome of this study). Mortality monitoring showed that over the 15-month observation period, there were nine deaths in the intervention group and 12 in the control group. Over the 27-month period, 23 and 35 people died in the intervention and the control groups, respectively. These differences were not statistically significant. The safety of the intervention was monitored by an independent committee, which conducted an interim analysis of the data every six months from the start of the study. Measures Outcomes Hospital utilization outcomes were quantified using Clalit’s hospitalization claims data, which include admission dates and type of admission (planned versus unplanned) in any general hospital. We calculated the number of unplanned admissions, the total number of inpatient days, and the number of 30-day readmissions during the first and second years of follow-up, excluding the first 3 months after enrollment to account for the initial assessment and patient on-boarding period41. Our original study protocol identified the primary outcomes as the percent change in admissions for ACSCs. However, due to the unavailability of data on the primary diagnosis at admission, we were unable to measure admissions specifically attributable to ACSCs as initially intended. Consequently, we revised our approach to focus on unplanned admissions more broadly, given that these are widely regarded as indicators of potentially preventable hospitalizations38. To reduce the risk of bias during data extraction, we systematically identified and extracted all admissions designated as “unplanned” in Clalit’s administrative and health data warehouse. These admissions included those occurring after an emergency department visit and those not pre-scheduled for a specific procedure, criteria chosen to provide a consistent, objective definition of unplanned hospitalizations rather than relying on potentially subjective interpretations of diagnostic characteristics. Patient-reported outcomes included perceived health and functional status. Patient health was assessed using the Physical and Mental Health scores of the 12-item Short-form (SF-12) Health Survey. Higher scores on the 0–100 Physical and Mental Health subscales represent better physical and mental health42. Functional status was assessed using the Katz Index of Independence in Activities of Daily Living (ADL) scale43 and a Brief Instrumental Activities of Daily Living (IADL) Measure44. Higher scores on both measures indicate greater difficulty in performing specified activities. The original study protocol identified another primary outcome as the expected reduction in total disease burden, based on the ACG® system score22. However, since the ACG® score is based on diagnoses, and the CC-MAP intervention often led to improved identification of conditions and additional diagnoses, changes in the ACG® score were determined to inaccurately reflect reductions in disease burden and thus were not included as an outcome in this study. Baseline characteristics A count of each patient’s chronic conditions was computed based on a list of 180 conditions in Clalit’s chronic disease registry. Clalit maintains a comprehensive database that integrates all Clalit patients’ ICD-9 codes (from primary care, specialty clinics, hospitals) and data from laboratories and pharmacies to compile a chronic disease registry. Designed for administrative and clinical management, the database is also available for research, and previous studies have demonstrated its validity45,46. Patient demographics were collected according to self-report on age, sex, education level (elementary school versus higher), living arrangement (living alone), and immigration status (immigrated since 1990 versus earlier immigration or native-born). Socioeconomic status (indicated by having a social security waiver for copayments, versus not having such a waiver) was collected from the Clalit administrated data systems. Throughout recruitment adequate representation of males and females was ensured. Sample size and statistical power We estimated the sample size required for the comparative-controlled trial on the basis of the trial’s primary outcome: unplanned hospital admissions. Our calculations were based on a previous 2-year cohort study of a similar primary care based, multimorbid care management intervention, which reported a mean increase of 0.18 in the rate of hospitalization in the control cohort (from 0.34 to 0.52) versus a non-significant decrease in the intervention group27. Accordingly, we assumed a modest effect size of 0.15, power = 0.90, and a two-sided α of 0.05. The sample size was calculated using G*Power 3.1 software47. Target sample size was 469. We increased this number to 600 in each arm to allow for testing the intervention with six fulltime equivalent nurses with a nurse-patient ratio of 1:00. Data analysis Our analyses of hospital utilization used an intention-to-treat approach. Intention-to-treat analysis mitigates bias associated with the exclusion of non-compliant participants, which can lead to exaggerated efficacy estimates, recognizing that noncompliance and protocol deviations are expected features of real-world clinical care48. Thus, all consented patients for whom hospital utilization data available were included in the analysis. We used univariate analyses (t-test for continuous variables and chi-square for binary variables) to explore differences between the study groups’ baseline characteristics (demographics and chronic morbidity) and outcomes (hospital utilization and patient-reported outcomes). To assess the degree to which variation was affected by the clustering of similar patients within clinics, we assessed the Intra Class Correlation (ICC) coefficient. ICC analysis showed weak clustering effects: ICC = 0.006 and 0.022 for unplanned admissions in years one and two, respectively. The ICC for the total inpatient days during years one and two was 0.010 and 0.024, respectively. Given these low coefficients, we did not employ random effects or nested-model designs, but rather we included clinic membership as a covariate in the adjusted models. ANCOVA analyses were used to adjust for identified differences in baseline characteristics in the final comparisons of the groups’ hospital utilization outcomes which were assessed in two periods - starting three months after enrollment and ending 15 and 27 months after enrollment, to allow for the time required for the nurses’ initial comprehensive assessments, care planning, and creation of the patients’ action plans. Blinding of the researchers was not feasible because the intervention and control arms differed in the number of participating clinics. This structural difference made group allocation readily identifiable within the dataset. Nonetheless, as outcomes and allocation were predetermined this did not in any way affect the analysis. Unplanned admissions were designated as the primary outcome, with total inpatient days and 30-day readmissions examined as closely related additional outcomes, reflecting complementary aspects of acute care utilization. No formal adjustment for multiple outcome testing was applied, consistent with methodological guidance indicating that such adjustments are not universally required when a limited number of clinically related outcomes are prespecified and interpreted in the context of the overall pattern of results49. One-year patient-reported outcomes were assessed in the subsample of participants who provided data for each outcome at both baseline and 1-year post-enrollment. For each model, participants were included if they completed at least 50% of items for each scale. Missing items were replaced with the item-level group mean for patient health and with the mean of the answered items for the functional status. As with the hospital utilization outcomes, ANCOVA was used to adjust for baseline group differences in participant characteristics. The first analyses were then repeated using the original (non-imputed) values, and they yielded comparable results. Data analysis was performed using R version 4.1.1. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting summary Transparent Peer Review file Source data Source Data Acknowledgements We would like to thank all of the Comprehensive Care for Multimorbid Adults Program (CC-MAP) nurses and the larger CC-MAP clinical and coordinating team and staff for their contribution to this study. Author contributions E.S. and R.D.B. conceptualized and designed the study; C.B. assisted with the conceptualization and adaptation of the study design; M.L. was responsible for data curation; O.T., M.L., and E.S. performed the data analysis; C.K., T.M., and S.G. were responsible for project administration and the acquisition of all study resources; C.K., D.N., and R.D.B. supervised all project phases; E.S. and O.T. drafted the paper. All authors critically reviewed and revised the paper for important intellectual content. Peer review Peer review information Nature Communications thanks Niall Grant and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Funding This work was partially supported by the Gertner Institute. The sponsors had no role in the study design, methods, subject recruitment, data collection, analysis, preparation of manuscripts, or the decision to publish. No additional external funding was provided. Data availability The raw data of this study are protected and are not available due to data privacy laws. In line with Clalit’s information governance policy the data cannot be shared in a repository or on an individual request basis without IRB approval, as specific consent for data sharing were not obtained. Specific requests for remote access to data should be referred to Clalit Research Institute. Subject to receipt of the aforementioned consent and subsequent approvals, data sharing will be made in a secure setting, on a per-case-specific manner, solely for the purpose of reproducing the analysis carried in the research paper, as defined by the chief information security officer of Clalit. Source data for Figs. 2 and 3 are provided with this paper. Source data are provided with this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-73104-1. References Associated Data Supplementary Materials Supplementary Information Reporting summary Transparent Peer Review file Source Data Data Availability Statement The raw data of this study are protected and are not available due to data privacy laws. In line with Clalit’s information governance policy the data cannot be shared in a repository or on an individual request basis without IRB approval, as specific consent for data sharing were not obtained. Specific requests for remote access to data should be referred to Clalit Research Institute. Subject to receipt of the aforementioned consent and subsequent approvals, data sharing will be made in a secure setting, on a per-case-specific manner, solely for the purpose of reproducing the analysis carried in the research paper, as defined by the chief information security officer of Clalit. Source data for Figs. 2 and 3 are provided with this paper. Source data are provided with this paper.