The absence of data leaves open the possibility that higher medication costs generated by physicians coding more mental health diagnoses offset lower inpatient and testing costs. If these physicians prescribed more antidepressants, particularly selective serotonin reuptake inhibitors, then the costs of these drugs may partially offset their lower expenditures in other areas. Total pharmacy expenditures for antidepressants, however, accounted for less than 2% of total expenditures in 1997, a percentage that has increased since 1995. Thus, it is unlikely that this missing information could explain the differences observed. The total expenditures include inpatient and outpatient mental health costs generated by each physician.
The use of risk of avoidable hospitalizations as a measure of appropriateness of care for biomedical problems is of questionable validity. Even if valid, it is a relatively insensitive measure. It is reassuring that the proportion of recorded mental health diagnoses was associated with lower avoidable hospitalization rates. This suggests that physicians with higher proportions of recorded mental health diagnoses are not neglecting biomedical problems in their patients and may be providing better medical care. However, one study found that a patient-centered or integrated approach (which addressed the psychosocial components of care) to non–insulin-dependent diabetes was associated with improved patient satisfaction and well-being but poorer control of cardiovascular risk factors.31 Further study on the relationship between physician style and patient outcomes is needed.
Our analyses are cross-sectional, so no conclusions about causality can be made. Unmeasured confounding may also account for the findings. The results may not be generalizable beyond the managed care organization studied. However, the subjects in our analysis represented more than 50% of the patients and 95% of the primary care physicians in the metropolitan area studied. In this managed care organization, there was no financial disincentive for making mental health diagnoses (they are reimbursed at the same rates as medical diagnoses), and 50% of the cost of outpatient specialist mental health treatment was covered.
We used diagnoses coded by the physicians for insurance billing purposes. Although coded diagnoses cannot tell us what occurs during the visit, we speculate that the coding of a mental health diagnosis requires that the physician recognizes the disorder and has some type of implicit or explicit negotiation with the patient that makes recording the diagnosis acceptable. It has been shown that physicians tend to underreport mental health diagnoses because of concerns about reimbursement or stigmatizing the patient.32 The proportion of recorded mental health diagnoses can be seen as a reflection of the extent to which the physician-patient interaction legitimizes this category of diagnoses.
Conclusions
There is increasing recognition of the importance of diagnosing mental health disorders in primary care and the significance of those diagnoses on patient outcomes and medical costs. The results of our study suggest that physicians who record higher proportions of mental health diagnoses have lower health care costs. These results are consistent with the hypothesis that increased detection and treatment of mental health disorders in primary care may reduce overall health care costs. Previous studies have demonstrated that psychosocial orientation and communication style can predict a physician’s ability to detect mental health problems.33-35 Physicians can be trained to improve their communication skills and detect and manage emotional distress.36 Randomized trials have demonstrated that communication training for physicians can result in improved detection of psychological disorders and less anxiety and emotional distress in their patients.37 Our results should provide further incentive to investigate how physicians’ practice styles differ and how to improve the primary care physician’s ability to recognize and diagnose mental health disorders.
Acknowledgments
This study was supported by a grant from the Agency for Policy Health Care and Research, R01 HS09397-01.