Background: Health systems must transition from catering primarily to acute conditions, to meet the increasing burden of chronic disease and multimorbidity. Case management is a popular method of integrating care, seeking to accomplish this goal. However, the intervention has shown limited effectiveness. We explore whether the effects of case management vary in patients with different types of multimorbidity. Methods: We extended a previously published quasi-experiment (difference-in-differences analysis) with 2049 propensity matched case management intervention patients, adding an additional interaction term to determine subgroup effects (difference-in-difference-in-differences) by different conceptualisations of multimorbidity: 1) Mental-physical comorbidity versus others; 2) 3+ chronic conditions versus < 3; 3) Discordant versus concordant conditions; 4) Cardiovascular/metabolic cluster conditions only versus others; 5) Mental health-associated cluster conditions only versus others; 6) Musculoskeletal disorder cluster conditions only versus others 7) Charlson index > 5 versus others. Outcome measures included a variety of secondary care utilisation and cost measures. Results: The majority of conceptualisations suggested little to no difference in effect between subgroups. Where results were significant, the vast majority of effect sizes identified in either direction were very small. The trend across the majority of the results appeared to show very slight increases of admissions with treatment for the most complex patients (highest risk). The exceptions to this, patients with a Charlson index > 5 may benefit slightly more from case management with decreased ACSC admissions (effect size (ES): -0.06) and inpatient re-admissions (30 days, ES: -0.05), and patients with only cardiovascular/metabolic cluster conditions may benefit slightly more with decreased inpatient non-elective admissions (ES: -0.12). Only the three significant estimates for the musculoskeletal disorder cluster met the minimum requirement for at least a small effect. Two of these estimates in particular were very large. This cluster represented only 0.5% of the total patients analysed, however, so is hugely vulnerable to the effects of outliers, and makes us very cautious of interpreting these as real effects.