Local, collaborative, stepped, and personalized care management for older people with chronic diseases – results from the randomized controlled LoChro-trial

In a two-group parallel RCT, LoChro-Care (intervention group, IG) was compared with usual care (control group, CG). The study was conducted at the Medical Center, University of Freiburg, Germany. Recruitment took place between January 2018 and March 2020. From the beginning of 2020, several procedures were adjusted due to the COVID-19 pandemic. From then on, we conducted the follow-up assessments, as well as the intervention contacts, exclusively by telephone. Ethical approval was granted by the ethics council of the University of Freiburg, Germany (495/17). The study follows the CONSORT guideline for reporting results from randomized trials [25] (see CONSORT Checklist – LoChro-study, additional file 1).

The a priori calculated sample size aimed for n = 606 study participants, assuming a dropout rate of 20% at the last follow-up time point (T₂), a significance level of 0.05, 95% power, and a standardized mean difference between IG and CG of 0.30. Recruitment took place at the emergency center, selective wards, and at the geriatric, diabetes, and memory outpatient clinics of the Medical Center, University of Freiburg. Thus, we recruited both inpatients and outpatients, who were enrolled by research associates. Eligible patients were older adults aged 65 years or above with one or multiple chronical illnesses or geriatric symptoms who lived in Freiburg and nearby surrounding areas. In total, n = 2,721 potential participants took part in a short screening using the German version of the “Identification of Seniors at Risk” (ISAR) screening tool [26] (Fig. 1, flow diagram), which assesses the participants’ risk of unplanned readmission and need for nursing care. The inclusion criterion for study participation was an ISAR total score of two or more. Exclusion criteria were an ISAR score of less than two, a terminal medical condition, and insufficient German language skills. From the eligible patients, n = 1,477 did not meet the inclusion criteria and n = 720 refused to participate because they were satisfied with their current care situation or due to a lack of interest or time, for example (Fig. 1). Thus, the study contained n = 524 participants in total, from which n_IG = 261 were randomly allocated to the IG receiving the LoChro-Care, and n_CG = 263 CG-participants remained under usual care. We applied a block randomization without stratification. The allocation to the two groups was done by a research associate who was not involved in the recruitment, data collection procedure, or intervention provision using a computer-generated randomization schedule.

Patients, or the legal representative in the case of severe dementia or comparable conditions, granted written informed consent for study participation. Afterwards, the participants took part in the first questionnaire survey (baseline, T₀). The follow-up questionnaire surveys took place 12 (T₁) and 18 (T₂) months later at the participants’ home (13.64% at T₁, 5.86% at T₂), if possible, or by telephone (86.36% at T₁, 94.14% at T₂). Responses were collected by research associates during individual interviews. In the case of medical conditions like dementia with limiting cognitive abilities for answering, the legal representative or relative responded to the questionnaire surveys (proxy-assessment). As our outcome variables primarily aimed at the subjective perception of health and psychosocial well-being, which can hardly be judged by third parties, these were not assessed via the proxy assessments. Thus, only the patients’ self-reported data are considered in the analysis reported here (patient assessment, see flow diagram Fig. 1).

LoChro-Care was designed as a local, collaborative, stepped, and personalized care intervention. It focuses on enhancing the patients’ self-management in coordinating their individual care network by providing assistance to maintain or establish contact to formal and informal support (e.g., general practitioner, family, regional geriatric outpatient services). It comprised 7–16 contacts with trained chronic care managers (CCM) and lasted 12 months. A total of four CCM with extensive experience in the sectors of nursing, health education, and social work provided the intervention (qualifications: CCM 1: Bachelor of Science (BSc) Nursing Science, nurse; CCM 2: Master of Arts (MA) Health Education, nurse; CCM 3: MA Social Work; CCM 4: BSc Health Education, MA Social Work student). At least the first three contacts were home visits. Depending on the constitution of the patient, the monitoring contacts could also take place by telephone. From the beginning of 2020, the intervention contacts were conducted exclusively by telephone due to the COVID-19 pandemic. Figure 2 shows the elements of the LoChro-Care intervention. Following the “Ariadne principles” for the management of multimorbidity in primary care [23], LoChro-Care included: a) a comprehensive assessment of the individuals’ health, psychosocial, and care conditions; b) the development of an individualized care plan in accordance with the patients’ prioritized health problems and preferences; c) the implementation, adaptation, and monitoring of the care plan; and d) a closing session. Extra modules were provided for patients with mild depression or diabetes, which were administered by the CCM responsible for the respective patient. In the case of mild depression, six additional contacts were implemented, including a short problem-solving therapy. In the case of diabetes, three extra contacts were included that aimed at the improvement of the patients’ self-management skills concerning diabetes. Additionally, there was an option to involve trained volunteers for support if patients could not implement the care plan on their own and no primary caregiver was available. For the implementation of the intervention, standardized action plans for each intervention contact between the CCM and the patient were developed. In addition, standardized care plans for recurring patient issues were created (incl. informational materials) and adapted to the individual patients’ care situation, considering formal and informal support. Given the case of sufficient contact with the general practitioner and family caregivers, the CCM informed the patient about self-management and local geriatric outpatient services suitable for the patient’s needs. In the case of insufficient support, the CCM assisted with establishing contact with formal and informal care providers, e.g., the general practitioner, primary caregivers, trained volunteers, or geriatric services. In the monitoring sessions, the CCM and the patient collaboratively evaluated the implementation of planned actions, the patient’s current health problems, needs, and preferences, and adapted the care plan accordingly. The CCM themselves neither implemented any specific treatment measures nor connected the patients directly to health or psychosocial services. Instead, they identified appropriate care providers and assisted the patients with contacting them. In order to ensure treatment fidelity, the CCM had regular supervision sessions by an interdisciplinary geriatric team and research associates responsible for the process evaluation, as well as team intervision.

The same measurements were collected in each group at each time point. The primary outcome was the patients’ physical, psychological, and social health status. This was represented by a composite score consisting of the participants’ functional health and depressive symptoms, as these two impairments are assumed to be highly prevalent in the target population and reciprocally associated [8, 9]. The first component, functional health, was assessed using the “WHO Disability Assessment Schedule 2.0” (WHODAS 2.0) [27]; the latter, depressive symptoms, was assessed with the “Patient Health Questionnaire” (PHQ-9) [28]. The WHODAS assesses aspects of functional health during the last 30 days via six domains: cognition, mobility, self-care, interactions with other people, activities of daily living, and social participation. For an overall WHODAS-score, responses of the 32 items were counted, and the resulting score was transformed to a scale range of 0–100. Higher scores thereby indicate lower levels of functional health. The PHQ-9 measures depressive symptoms according to the DSM-IV criteria (Diagnostic and Statistical Manual of Mental Disorders) and consists of nine items (reference period: the last two weeks). The summary score ranges from 0–27, with a higher score indicating more severe depressive symptoms. In order to bring these differently measured outcomes to the same metric and, thus, to calculate the composite score, the Proportion-of-Maximum-Scaling (POMS) method was used [29]. Therein, each scale was transformed into a scale ranging from 0 (minimum) to 100 (maximum) using the following formula:

One secondary outcome was the patient’s evaluation of their health care situation. This was assessed by the “Patient Assessment of Chronic Illness Care” (PACIC) [30] (German version), which asks for the patients’ experiences with the described care elements within the last 6 months (11 items) and their overall satisfaction with health care (one item). For the analyses presented here, we used the one item of interest, namely: “In what percentage of cases was I satisfied with the organization of my medical care”. The two other secondary outcomes were health-related quality of life (HRQL) (using a self-developed item) and life-satisfaction (LS) [31]. Each of these was assessed with a single-item scale ranging from 0–10, with higher values indicating a higher HRQL or LS, respectively.

Multimorbidity was assessed using a weighted multimorbidity index, which can have values from zero to 37 [32]. Information on multimorbidity stemmed from the hospital’s documentation system and additional patient-reported diagnoses.

All statistical analyses were executed using SPSS (version 28). Both intention-to-treat (ITT) and per-protocol (PP) analysis were performed. The statistician was blinded to the group allocation. An alpha-level of 0.05 (two-sided) indicated statistical significance.

The analysis compared change over time within the outcome variables between both groups. Linear mixed modelling (LMM) analysis was applied due to its robustness in handling missing data, as well as being able to handle both within- and between-subjects effects [33]. Intra-individual change over time was modelled on the first LMM level, while the subject-related inter-individual comparison was modelled on the second level [34]. For each primary and secondary outcome, a LMM was built following three consecutive steps: first, a random-intercept model (M₁) for inter-class correlation (ICC) estimations was constructed including a random intercept for each participant. Second, a loaded model (M₂), including all of the fixed effects with a respective choice of covariance matrix, was modelled. In a third step, the full model (M₃) included the repeated measurement effect and defined covariance matrix [33]. The model fit was compared using likelihood ratio tests (LRT), based on the respective -2 log-likelihood indices and degrees of freedom per model [35].

Control variables were included since adjusted analyses are also recommend as sensible for RCTs when certain variables are expected to be prognostic [25]. Hence, the control variables comprised individual characteristics, namely gender, age, degree of multimorbidity, family status, residential status, health insurance status, and educational level, as well as self-reported depression and diabetes at baseline, as these two groups received specific LoChro-Care components. Additionally, differences in recruitment path (e.g., inpatient and outpatient setting) and interview setting (face-to-face vs. telephone) were included. The categorical control variables were dummy coded. The missing data and sample-dropouts were handled through the application of maximum likelihood estimation (restricted). As a long-format of the data set was used, the baseline values of each variable, as well as the values at T₁ and T₂, are represented through the combination with the measurement time point variable [33].

Systematic differences between dropouts and non-dropouts were examined with regards to baseline characteristics using t-tests for independent samples and contingency tables.

At baseline, a total number of N = 501 participants were assessed (n_IG = 249 and n_CG = 252; flow diagram Fig. 1 “patient assessment”). The dropout rate was 32.34% (n = 162) after 12 months (T₁) and 7.96% (n = 27) after 18 months (T₂). Dropouts showed more impairments and depressive symptoms, and higher levels of multimorbidity. Unplanned inpatient hospitalizations (e.g., because of falls or stroke) were documented as Serious Adverse Events (SAE) in n = 28 cases, but we did not exclude these cases from our study. Exclusion criteria for the analysis were: a) pilot phase participation, b) more than 50% missing values, c) cases where the respondent changed from proxy to patient in comparison to baseline, and d) proxy assessment. The ITT sample consisted of n_ITT = 491 participants. IG-participants who did not finish the intervention with a closing session were excluded in the PP analysis. Among these excluded IG-participants, n = 54 (20.69%) did not receive LoChro-Care (e.g., because of death [n = 15, 27.78%] or decline in the health condition [n = 6, 11.11%]; Fig. 1). Additionally, n = 37 (14.18%) IG-participants discontinued the intervention. Thus, the PP sample consisted of n_PP = 408 cases.

At baseline, the mean age in the ITT sample was M = 76.78 (SD = 6.35) years, ranging from 65 to 94 years. In total, n = 275 (56.01%) participants were female, n = 229 (47.64%) were married, and n = 448 (91.75%) lived in their private homes. The multimorbidity-index mean was M = 5.46 (SD = 3.44), ranging from zero to 20. Further demographic characteristics of the ITT, as well as the PP sample, can be found in Table 1. The baseline means and standard deviations for all outcomes can be found in the supplementary Table S1 (additional file 2).

For the composite score (ITT), the ICC with r = 0.54 was relatively high due to the nature of the repeated measurement design. The model specifications (loaded model and full model) significantly enhanced the model-fit (Χ₁(21) = 438.61, p < 0.001; Χ₂(2) = 6.36, p = 0.02) [36]. Further information on the model fit and likelihood-ratio tests for the LMM of each outcome can be found in Table S2 (additional file 3).

Although we reached a comprehensive sample size in the context of geriatric research, which formed a sound basis for the statistical analyses, some limitations should be mentioned. Potential bias could result from the regional specificity and the exclusion criteria applied. The study area was restricted to Freiburg and surrounding areas. Specific characteristics of this area, like the relatively high socioeconomic performance, might have influenced the implementation of the intervention and study results. Therefore, future studies should investigate similar health care approaches for older people with multiple chronic diseases in other German areas for comparison. In addition, we did not include patients with terminal conditions or insufficient German language skills. These factors could likely be conditions occurring in the population of older multimorbid people and in ethnically diverse societies, which pose specific demands on the care management; therefore, they should be explored in future research.

Finally, the effect of the COVID-19 pandemic cannot be ruled out. From the beginning of 2020, several study procedures were adapted to the pandemic situation. The monitoring and closing sessions of the intervention were then primarily provided by telephone. In principle, this was not an issue of particular concern because the intervention design and manual had already included telephone contacts between the CCM and patients. Moreover, the telephone contacts were feasible in most cases. Nevertheless, the general negative effects of the COVID-19 pandemic, such as restricted contacts, reduced doctor visits, or impaired mood, could have interfered with the intervention and influenced the evaluation of its effectiveness.

In sum, we can infer several practical implications for future research and practice as indicated above. Attempts to modify LoChro-Care or to develop new interventions for older multimorbid people could include more active assistance in establishing formal and informal supports. In addition, it could comprise comprehensive case management that goes beyond self-management support. In this regard, the patients’ degree of multimorbidity, severity of the already existing health conditions, and prognostic progression should always be considered, and the optimal time-point for treatment initiation needs to be determined. In addition, it would be worthwhile to give more attention to both specific diagnoses with potential impact on the health status (e.g., depression) and sociodemographic factors. In particular, gender-specific needs could be addressed in the provision of care. In this context, it could be beneficial to gain more insight into the patients’ individual needs and perceived helpful intervention components, as well as into the care managers’ perspective on feasible care elements. Therefore, future research could first use qualitative methods (e.g., interviews) to explore the demanded intervention elements. Such results, in turn, might inform the development of care interventions and facilitate the identification of potential effective care management elements in respect to the target group and intervention goal.