There are many limitations to this analysis. All study patients had access to a psychiatrist and were taking medication. Furthermore, the very process of daily recording may improve medication adherence (van Berge Henegouwen et al. 1999). In clinical practice, most patients would not volunteer to keep a long-term daily record. This study may be less generalizable to other patients with bipolar disorder. With a naturalistic design, the patients in this study varied in severity, phase and course of the disease, and years of illness. The study included more females than males and more patients with bipolar I than bipolar II disorder. All data were self-reported and daily access to a computer was required. The strengths of this study include the use of prospective data from a relatively large number of patients and the entry of specific drugs and dose taken daily. However, the prescribed regimens and reasons for addition of new drugs were not known. Some drugs may be misclassified as to prescribing intent. For example, trazodone is considered an antidepressant but is frequently prescribed as a hypnotic agent. The analyses did not include other classes of psychotropic drugs such as stimulants, drugs taken for medical conditions, and over-the-counter preparations. Finally, many diverse factors that may influence medication selection such as physician prescribing habits, drug costs, regulatory influences, insurance plan formularies, pharmaceutical marketing, and patient preferences were not considered (Thistlewaite et al. 2010; Poulsen 1992).