I don't know about SPSS as I don't use the program but you can run mediation models with binary/dichotomous outcome variables using the "mediation" package (https://cran.r-project.org/web/packages/mediation/vignettes/mediation.pdf).

For your variables, you can combine the "lm" and "glm" functions to derive two different regression models (one for OLS linear regression with the mediator as the outcome variable, and one for binary logit regression with the dichotomous variable as the outcome) and combine them in the mediation package to estimate the indirect effect.

Otherwise, you can also run it as a single model using SEM through the "lavaan" package where you can designate the outcome as being binary (https://lavaan.ugent.be/tutorial/cat.html).

You should be able to run this analysis using Hayes’ PROCESS macro, which runs in SPSS. You’ll likely use Model 4. Here’s a link to the macro:

https://www.processmacro.org/index.html

Sea cual sea la mirada que se le da a la escala lo importante es que representa una variable de percepción cualitativa por lo que se requiere de un tratamiento estadístico alejado de resultados de tipo paramétrico o meramente descriptivo, lo que justifica el uso de la técnica de correspondencias que busca

l criterio que se siguió para interpretar los resultados de la encuesta con una escala de Likert de 5 puntos fue: Respuestas mayores o iguales a 3.5 son consideradas como aceptables, y respuestas menores a 3.5 se consideraron como no aceptables.

Karlson, K. B., Holm, A., & Breen, R. (2012). Comparing regression coefficients between same-sample nested models using logit and probit : A new method. Sociological methodology, 42(1), 286-313.