ObjectiveTo ascertain the probable validity of a comprehensive complexity care model in the management and referral of complex cases by studying the agreement between complexities observed from different levels of care in complex chronic patients, advanced chronic patients and at the end of life, using this model as a tool. MethodsSource: PADES database (demographic and descriptive data) and opinion of the participants. Participants: the 3 referring (participating) professionals of each case (Primary Health Care [APS], Palliative Care Support Team [PADES], Specialist [ESP]). Variables: age, sex, patient typology (oncological, non-oncological), evolutionary or prognostic classification (chronic complex patient [PCC, advanced [PCA], end of life [PFV]). Model application: 1) detected complexity: clinical, psychoemotional, socio-familial, spiritual, ethical, and related to death; 2) degree of detected complexity: the highest complexity level achieved in the evaluated areas and 3) PADES intervention proposal: low complexity: occasional attention, medium: shared caring; high complexity: intense attention). Complexity detected and intensity of PADES intervention in each case according to APS, PADES, ESP: agreement analysis. Study: transversal observational in 2 phases (prospective and retrospective). Statisticians: Cohen's kappa index and χ2 (kappa comparison). Analysis: IBM SPSS v. 23 and Epidat v. 4.2. ResultsThe study encompassed 500 patients —prospective phase 248 and retrospective 252—, 54.3% women and 64.8%≥80 years. A 26.8% oncological, and 51.8% PCC, 26.4% PCA and 21.8% PFV. Complexity detected: medium/high, with an average of 3 areas affected by case (mainly clinical, psychoemotional, socio-familial, ethical). Agreement in the intervention intensity PADES with APS and ESP: kappa indices are good or very good (>0.80) at a global level, with both APS (kappa 0.92, 95% confidence interval: 0.89-0.95) as with ESP (kappa 0.83, 95% confidence interval: 0.78-0.87). This high agreement even improves with APS in the retrospective phase and is also maintained and is homogeneous when analyzed according to the typology and the prognostic classification of the patient. Only with ESP are there differences between categories, with P<0,0001 for all kappa presented. ConclusionThe highest agreement among observers in relation to the degree of support teams’ intervention according to the level of detected complexity suggests that this explored complexity model is adequate for the management and referral of complex cases, irrespective of the clinical typology (oncological or otherwise) and the patient's prognosis.