Using qualitative synthesis to explore heterogeneity of complex interventions

Background: Including qualitative evidence on patients' perspectives in systematic reviews of complex interventions may reveal reasons for variation in trial findings. This is particularly the case when the intervention is for a long-term disease, as management may rely heavily on the efforts of the patient. Inclusion though seldom happens, possibly because of methodological challenges, and when it does occur the different forms of evidence are often kept separate. To explore heterogeneity in trial findings, we tested a novel approach to integrate qualitative review evidence on patients' perspectives with evidence from a Cochrane systematic review. Methods: We used, as a framework for a matrix, evidence from a qualitative review on patients' perspectives on helping them manage their disease. We then logged in the matrix whether the interventions identified in a Cochrane review corresponded with the patient perspectives on how to help them. We then explored correspondence. The Cochrane review we used included 19 trials of interventions to improve adherence to therapy in HIV/AIDS patients. The qualitative review we used included 23 studies on HIV/AIDS patients' perspectives on adherence; it translated the themes identified across the studies into recommendations in how to help patients adhere. Both reviews assessed quality. In the qualitative review they found no difference in findings between the better quality studies and the weaker ones. In the Cochrane review they were unable to explore the impact of quality in subgroup analysis because so few studies were of good quality. Results: Matrix tabulation of interventions and patients' perspectives identified a range of priorities raised by people infected with HIV-1 that were not addressed in evaluated interventions. Tabulation of the more robust trials revealed that interventions that significantly improved adherence contained more components considered important by patients than interventions where no statistically significant effect was found. Conclusions: This simple approach breaks new ground in cross tabulating qualitative evidence with the characteristics of trialled interventions. In doing so it tests the assumption that patients are more likely to adhere to interventions that match more closely with their concerns. The potential of this approach in exploring varying content and rates of success in trialled complex interventions deserves further evaluation.