Review of Statistical Conclusions of National Trends in Psychotherapy by Office-Based Psychiatrists

Implications: An August 2008 study analyzing the National Ambulatory Medical Care Survey (NAMCA) database shows a significant decrease in the amount of psychotherapy that psychiatrists are providing for their patients, while at the same time possibly showing an increase in medications prescribed. This correlation implies an emerging trend away from the traditional role of the psychiatrist as providing therapy to control common psychiatric disorders. We looked at the study design and statistical analyses to determine what impact this study should have as a commentary on possible changes in prescription rates.

Analysis: In this study, NAMCA survey data ranging from 1996 to 2005 was analyzed. They used a volunteer population of non-federally employed, office-based physicians generally involved in direct patient care.  Systematic random samples of visits to each physician were drawn for a randomly selected week.  

According to the study, 44.4% of visits to psychiatrists in 1996-1997 included psychotherapy, in comparison with 28.9% between 2004-2005. The authors attempted to correlate with this decrease in psychotherapy an increase in prescriptions which may not be a valid conclusion.  

Issues with the study design include a) a relatively low survey response rate (61,5%) from ‘volunteer’ psychiatrists which could introduce significant bias; b) survey did not take into account patients who used dual-therapy involving a non-physician mental health professional which might result in the reduction of psychiatrist psychotherapy, and c) no analysis was conducted to directly assess changes in prescription rates.  

Table 1 odds ratios do not relate to any change in prescription but rather relate to the odds of receiving psychotherapy for patients who did receive prescriptions (OR = 0.93) or did not receive medication (OR = 0.90).  One may be tempted to use the raw numbers of prescriptions for comparison and conclude that in 1996-1997 the prescription rate waslower at 1808/2637 = 0.685 then 2841/3389 = 0.838 in years 2004-2005 (P < 0.0001).  However, this is incorrect for two reasons.  First, these raw rates are incorrect since they are not weighted appropriately using the NAMCS sampling weights.  They cannot be used with any known accuracy.  Second, since multiple patients are seen by a single psychiatrist, these raw values have not been adjusted for the variability in provider prescription rates.  A mixed model would be required to appropriately analyze these data.  

Table 2 seems to suggest the inverse relationship of medication use with psychotherapy (OR = 0.25) can be interpreted as evidence of an increase in prescriptions.  While this interpretation is possible, several others could be derived to explain these results.  For example, if psychiatrists replaced medications with psychotherapy the same inverse relationship would be seen without an increase in prescription rate, and in fact could be seen with a decrease in prescription rate.   Thus, multiple hypotheses might exist for this relationship with no way to decide which hypothesis is correct among them.   

Although the authors comment that a correlation between the number of psychiatrists providing psychotherapy to their patients and amount of prescriptions issued may be occurring, there are too many outstanding factors and data analysis discrepancies to claim that the decrease in psychiatrists providing psychotherapy directly results in a rise in the amount of psychopharmacological treatments being prescribed.  

For further information about our assessment of these and other studies, please contact your GLG representative to schedule an individual consult.  

Research and initial writing of this article was done by Ms. Alexandra McWilliams, BioRankings LLC.