Our search identified 654 unique papers, which after initial assessment (CW) were reduced to 25. These remaining 25 were independently scrutinised for inclusion by PV, MT and CW. When the reviewers did not agree, the paper was discussed and consensus reached. Finally, seven papers were included. Figure 1 illustrates the "flow" of papers.
The meticulous review of each included paper is reported in Tables 1, 2, and 3 (see additional file 1, 2, and 3). CW primarily filled in the Tables, and MT, PV and CW all read the included articles and checked the contents of the tables.
Characteristic elements of interventions are outlined in Table 1 (Consort items 1–5). Outcomes of interest, statistical methods, baseline data, number of participants analyzed, and effects found (Consort items 6, 8, 12, 15, 16, 17, and 18) are presented in Table 2. Table 3 presents elements of study validity found to be important: sample size, recruitment, random allocation, blinding of assessor, participant flow (Consort items 7, 14, 9, 11, and 13). Aspects of generalisability to non-participants are reported in second column points c) and d).
Main findings
Table 1 audits the characteristics of the seven included studies. Six trials were conducted in the USA, one in the UK. Only two of the six papers termed the intervention "CM" [19, 22]. The five other studies were deemed to fulfil reviewers' inclusion criteria (viz. the quoted CM definition). Interventions were classified: "advanced practice nursing" [23, 24], "home care interventions" (two were tested against usual care in a three armed trial [25], "care coordination" [26], and "nurse-led follow-up" [27]. Despite the different naming, all interventions will be designated "CM" throughout the rest of this paper.
Two studies [19, 23] included breast cancer patients only, two studies lung cancer patients only [25, 27]. The last three included different cancer types of which one trial [26] also addressed other advanced illness (chronic obstructive pulmonary disease (COPD) and chronic heart failure (CHF)). Two studies tested hospital-to-community interventions [19, 23], and three studies tested community interventions [22, 24, 25]. One study [27] tested a hospital (in-patient) intervention. The setting for one study was unclear, but the text implies an in-patient setting [26].
Health care professionals performed all interventions and nurses exclusively performed the intervention in five trials [19, 22–24, 27]. In one study, a nurse education was not a prerequisite to performing the intervention [26]. In another study [25] (the three-armed RCT), nurses were members of the intervention teams which included various professions.
Six studies described some sort of model, manual and/or use of supportive tools, e.g. assessment tool or checklist (Table 2). The sixth study [25] reported no such use of manual or tools, and offered no precise description of intended and performed intervention.
The patients' GPs were only expressly mentioned in the study from the UK [27]. This study also evaluated the intervention in terms of the GPs' satisfaction.
Table 2 presents outcomes studies. First of all, it is noteworthy that only three articles [19, 24, 27] mentioned which outcomes were primary and which were secondary. Since there is no overlap of outcomes studied (tool and methods of assessment) between the trials, no synthesis of effects can be made.
To further evaluate the effects, outcomes were divided (our own categories) into objective and subjective (= patient reported) elements. Objective elements studied included: survival [24, 27], received therapy [19], physical function (arm function) [19], advance directives and "do-not-resuscitate and intubate"-wishes [26], service use and hospitalization [22, 25, 27], and costs [23, 26, 27]. Subjective elements studied included: patient-reported needs and symptoms [22, 25], patient-reported "quality of life" [22, 24, 27], patient satisfaction and evaluation of the decision-making process [19, 26, 27], and relative-reported problems experienced in the interaction with the health care system [26].
When categorising outcomes according to the above criteria and taking nothing else into account, some dimensions of Quality of Life appeared to be improved. Similarly, all three papers reporting patient satisfaction reported improvements.
Looking at Table 3, it is obvious that validity elements of interventions were inadequately reported in most papers. Sample size measures were only reported in one paper [27], and one or more of the following elements were only poorly described: recruitment, allocation concealment method and blinding of assessor. Moreover, patient flows were incomprehensible in all papers except for one [27], which also included a flow diagram. All articles analysed whether randomization was successful regarding baseline covariates (Table 2).