Personal Health Records (PHRs), integrated with data from various sources, such as social care data, Electronic Health Record data and genetic information, are envisaged as having a pivotal role in transforming healthcare. These data, lumped under the term 'big data', are usually complex, noisy, heterogeneous, longitudinal and voluminous thus prohibiting their meaningful use by clinicians. Deriving value from these data requires the utilization of innovative data analysis techniques, which, however, may be hindered due to potential security and privacy breaches that may arise from improper release of personal health information. This paper presents a HIPAA-compliant machine learning framework that enables privacy-preserving classification of next-generation PHR data. The predictive models acquired can act as supporting tools to clinical practice by enabling more effective prevention, diagnosis and treatment of new incidents. The proposed framework has a huge potential for complementing medical staff expertise as it outperforms the manual inspection of PHR data while protecting patient privacy.