Depression Diagnoses and Antidepressant Use in Primary Care Practices

Primary care clinicians, who by most estimates provide the majority of mental health care in the United States, are often criticized for the underdiagnosis and undertreatment of patients with depression.¹ However, most studies antedate the widely publicized 1993 Agency for Health Care Policy and Research (AHCPR) primary care practice guidelines.^2,3 A recent survey of 1350 primary care physicians documented diagnostic and treatment patterns consistent with high-quality care,⁴ but more information about actual practice is needed. Studies that examine depression management in the context of primary care using tools that enable the assessment of episodes of care are needed.⁵

The purpose of our study was to examine the initial pharmacologic management and follow-up of patients with newly diagnosed depression in primary care practice. The use of antidepressants among patients not diagnosed with depression was also examined. Our study was done within the Practice Partner Research Network (PPRNet), a national practice-based network of primary care physicians who use a common electronic medical record (Practice Partner, PMSI, Seattle, Wash) and pool data for research.⁶

Methods

We used a retrospective cohort design with the calendar year 1996 as the primary period of observation. Each month participating practices run a computer program to extract patient-level data from their electronic medical records. The first time the program runs, all existing data are extracted; in subsequent months only new data are extracted. To protect patient confidentiality, the program assigns a unique anonymous numerical identifier to each patient. Extracted data include diagnoses, medications, vital signs, laboratory results, and other variables; the text from progress notes and reports are not extracted. The information is copied to diskettes and mailed to the PPRNet coordinating center. At the center, data are aggregated and undergo rigorous quality control and bridging of text strings of diagnoses and medications to standard data dictionaries. Original text strings of diagnoses are maintained in the database. Updated data tapes are sent quarterly to the PPRNet offices at the Medical University of South Carolina Center for Health Care Research, where they are converted to SAS data sets (Statistical Analysis System, Cary, NC) on a UNIX computer. PPRNet staff use microcomputers with standard database and statistical software to perform all subsequent analyses.⁷

Practice eligibility criteria for our study included family medicine or general internal medicine specialty, enrollment in PPRNet before July 1, 1995, and reliable recording of diagnoses and medications. Patients were eligible for the study if they had at least one entry in their electronic medical records during the years 1993-1995 and were at least 18 years old by December 31, 1996.

We identified depression in patients with text strings on the problem list that bridged to depression diagnoses. Patients with one of these diagnoses or a prescription for an antidepressant medication dated earlier than January 1, 1996, were excluded from treatment analyses. We calculated incidence rates for depression as the number of new diagnoses divided by the number of eligible patients.

We classified psychopharmacologic agents as either antidepressants or other psychotropic medications. We subclassified antidepressants as (1) selective serotonin reuptake inhibitors (SSRIs); (2) tricyclic and tetracyclic agents (TTAs); or (3) other antidepressants. Other psychotropic medications were subclassified as (1) benzodiazepines; (2) other antianxiety agents; or (3) others, which included antipsychotic and antimania medications.

For new cases of depression, we defined a “definite follow-up” as another contact where a diagnosis of depression or another prescription for an antidepressant medication was recorded. Since physicians often address problems beyond those documented at each visit, we defined a “possible follow-up” as all definite depression follow-ups plus contacts for other diagnoses.

Initial analyses revealed that a large proportion of patients who received antidepressants in 1996 had never been diagnosed with depression. To explore this issue, we examined all diagnoses for these patients within 5 days of the date of the antidepressant prescription. We performed analyses for all eligible patients in PPRNet and for each practice. Results are reported using simple descriptive statistics. We used chi-square statistics for comparisons of nominal data.

Results

Demographic Information

Thirty-nine of the 68 PPRNet practices were eligible for the study. The 29 noneligible practices were excluded for the following reasons: enrollment in PPRNet after July 1, 1995 (21); specialty or urgent care practices (7); or did not reliably record diagnoses (1). Sixteen (41%) of the eligible practices were in rural environments; the remainder were in suburban or urban settings. The specialty distribution among the eligible practices included 32 family medicine (7 academic, 25 community), 6 general internal medicine, and 1 combined family medicine and internal medicine. A total of 389 physicians saw patients in the 39 practices during the study year.

There were 149,327 patients eligible for the study, 88,727 women (59.4%) and 60,272 men (40.4%). The gender was unknown for 328 (0.2%) patients. The age distribution was as follows: 54.7% aged 18 to 45 years; 26.0% aged 46 to 65 years; and 19.3 % aged 66 years or older. The median number of patients per practice was 2785 (range = 192 - 21,120). Socioeconomic information, such as occupation or insurance coverage, is not consistently maintained in the PPRNet database.

Diagnosis and Treatment of Depression

We excluded 18,186 (12.2%) of the 149,327 patients from this portion of the analysis. Among those excluded, 3231 (17.8%) had a previous diagnosis of depression, 7658 (42.1%) had previously received an antidepressant, and 7297 (40.1%) had been given a depression diagnosis and an antidepressant prescription before 1996. A greater number and proportion of women (13,329, 15.0%) than men (4852, 8.0%) were excluded.

A total of 2103 (1.6%) of the 131,141 patients without a documented history of depression or antidepressant prescription had a new diagnosis of depression in 1996 Table 1. There were 213 problem text strings in the electronic medical record that were bridged to depression for these 2103 patients. However, 72% were categorized as “depression,” and 6% were “depressive disorder.” Women were more likely than men to receive these diagnoses. Wide variation was found among the 39 practices in the incidence of depression (range = 0.4% - 4.0%).

Of the newly diagnosed patients, 1032 (49.1%) were started on antidepressants within 5 days of the initial visit for depression Table 1. Of those, 934 were given antidepressants alone, and 98 patients were also prescribed other psychotropic medications. Ninety-four newly diagnosed patients (4.5%) were prescribed only nonantidepressant psychotropic medication, and 977 (46.5%) patients not prescribed psychotropic medications within 5 days. Of the antidepressants prescribed within 5 days of depression diagnoses, 81% were SSRIs and 13% were TTAs. Of the other psychotropic medications prescribed, 73% were benzodiazepines and 14% were other antianxiety medications. Three hundred fifty-three (33.0%) of the 1071 patients diagnosed with depression who were not prescribed antidepressants within 5 days of their diagnoses received an antidepressant prescription by the end of 1996.

Among the 1032 patients given antidepressants, 69.9% had at least one definite follow-up appointment, and 90.5% had at least one least one definite or possible follow-up contact in the 6 months after their diagnosis. Among those with a follow-up, the mean number of contacts was 5.3.

Total Antidepressant Use

Overall, 9335 (6.3%) of the 149,327 patients received a prescription for an antidepressant in 1996 Table 2. Wide interpractice variability was present in the prescription of antidepressants (range = 1.9% - 13.6%). A total of 6443 diagnoses were made for these patients within 5 days of their antidepressant prescription. We present the distribution of these diagnoses in Table 3. One third of the patients (3045) who received antidepressants had a previous diagnosis of depression and had received antidepressants before 1996. Those patients were considered prevalent cases of depression and account for 2% of the total population. Four thousand twenty-two patients (more than 40% of the patients who received an antidepressant in 1996) never received a diagnosis of depression.

Discussion

The strength of our study was our ability to examine a relatively large number of primary care practices and patients, and to explore, in a longitudinal manner, an important mental health care issue in primary care. With appropriate caveats concerning the generalizability of findings from physicians who choose to use electronic medical records, this study adds additional information to the field’s growing body of knowledge.⁵

Our study confirmed that depression management and antidepressant prescribing are important aspects of primary care practice. During 1996, 1 of every 48 women and 1 of every 104 men who were patients in PPRNet practices received a new diagnosis of depression. Of those diagnosed with depression, almost half received an antidepressant prescription within 5 days of the diagnosis. By the end of 1996, nearly two thirds of these patients had received such a prescription, a proportion similar to that found in the standard care arm of a primary care depression trial.⁸ The vast majority of treated patients received SSRIs, a finding consistent with the increasing use of these agents among outpatient psychiatrists.⁹ SSRI antidepressants are better tolerated than older agents and do not increase overall treatment costs,¹⁰ justifying their widespread use.

Nine of every 10 patients treated with antidepressants had at least one follow-up contact with the practice. On average, these patients had more than 5 contacts in the 6 months immediately following the diagnosis. This degree of follow-up is consistent with the AHCPR guideline, which recommends contact every 10 to 14 days for the first 6 to 8 weeks of treatment and every 4 to 12 weeks after that.³ This finding is in contrast with the concern that this degree of follow-up is not provided in primary care settings.¹

Obscured in the overall analyses are the nearly 10-fold interpractice variations in the diagnosis of depression (0.8% to 8.6% for both 1996 and earlier diagnoses) and the use of antidepressants (1.9% to 13.6%). Patient-level factors may account for some of this variation. It is likely, though, that there are physician-level differences in the ability to recognize, the willingness to diagnose, and the comfort in the treatment of depression.^1,11 Excluding minor depression and dysthymia, which may be as prevalent, epidemiologic studies have found a prevalence of depression from 6% to 8% in primary care samples.¹² Only 4 of the 39 PPRNet practices in our study had a 6% or greater prevalence of depression. However, 21 practices prescribed antidepressants to 6% or more of their patients. Since many of the diagnoses for these patients are not accepted indications for antidepressants, it may be that physicians are treating more depression than they are recording. Nonetheless, it is clear that opportunities remain for improvement in depression recognition among nearly half of the study practices.

Limitations

Several limitations to our study are important. Diagnoses listed in the electronic medical record reflect physician opinion and are not validated against a gold standard. The PPRNet database has limited information about nonpharmacologic treatment, so the extent to which patients received counseling or watchful waiting is unknown. Both of these strategies are endorsed by the AHCPR guidelines for patients with mild depression.³ The nature of the database also makes it difficult to be certain about the duration of pharmacotherapy or patient compliance with prescribed medication. Finally, it is uncertain whether clinicians actually addressed the patient’s depression during follow-up contacts.

Conclusions

Our study reveals the interpractice variations of diagnosing depression and prescribing antidepressants in primary care. Patients with newly diagnosed depression were followed up in a manner consistent with published guidelines. Opportunities for increased detection and treatment of depression exist in approximately half of the study practices. The unique features of PPRNet provide opportunities for future work in this area.

Primary care clinicians, who by most estimates provide the majority of mental health care in the United States, are often criticized for the underdiagnosis and undertreatment of patients with depression.¹ However, most studies antedate the widely publicized 1993 Agency for Health Care Policy and Research (AHCPR) primary care practice guidelines.^2,3 A recent survey of 1350 primary care physicians documented diagnostic and treatment patterns consistent with high-quality care,⁴ but more information about actual practice is needed. Studies that examine depression management in the context of primary care using tools that enable the assessment of episodes of care are needed.⁵

The purpose of our study was to examine the initial pharmacologic management and follow-up of patients with newly diagnosed depression in primary care practice. The use of antidepressants among patients not diagnosed with depression was also examined. Our study was done within the Practice Partner Research Network (PPRNet), a national practice-based network of primary care physicians who use a common electronic medical record (Practice Partner, PMSI, Seattle, Wash) and pool data for research.⁶

Methods

We used a retrospective cohort design with the calendar year 1996 as the primary period of observation. Each month participating practices run a computer program to extract patient-level data from their electronic medical records. The first time the program runs, all existing data are extracted; in subsequent months only new data are extracted. To protect patient confidentiality, the program assigns a unique anonymous numerical identifier to each patient. Extracted data include diagnoses, medications, vital signs, laboratory results, and other variables; the text from progress notes and reports are not extracted. The information is copied to diskettes and mailed to the PPRNet coordinating center. At the center, data are aggregated and undergo rigorous quality control and bridging of text strings of diagnoses and medications to standard data dictionaries. Original text strings of diagnoses are maintained in the database. Updated data tapes are sent quarterly to the PPRNet offices at the Medical University of South Carolina Center for Health Care Research, where they are converted to SAS data sets (Statistical Analysis System, Cary, NC) on a UNIX computer. PPRNet staff use microcomputers with standard database and statistical software to perform all subsequent analyses.⁷

Practice eligibility criteria for our study included family medicine or general internal medicine specialty, enrollment in PPRNet before July 1, 1995, and reliable recording of diagnoses and medications. Patients were eligible for the study if they had at least one entry in their electronic medical records during the years 1993-1995 and were at least 18 years old by December 31, 1996.

We identified depression in patients with text strings on the problem list that bridged to depression diagnoses. Patients with one of these diagnoses or a prescription for an antidepressant medication dated earlier than January 1, 1996, were excluded from treatment analyses. We calculated incidence rates for depression as the number of new diagnoses divided by the number of eligible patients.

We classified psychopharmacologic agents as either antidepressants or other psychotropic medications. We subclassified antidepressants as (1) selective serotonin reuptake inhibitors (SSRIs); (2) tricyclic and tetracyclic agents (TTAs); or (3) other antidepressants. Other psychotropic medications were subclassified as (1) benzodiazepines; (2) other antianxiety agents; or (3) others, which included antipsychotic and antimania medications.

For new cases of depression, we defined a “definite follow-up” as another contact where a diagnosis of depression or another prescription for an antidepressant medication was recorded. Since physicians often address problems beyond those documented at each visit, we defined a “possible follow-up” as all definite depression follow-ups plus contacts for other diagnoses.

Initial analyses revealed that a large proportion of patients who received antidepressants in 1996 had never been diagnosed with depression. To explore this issue, we examined all diagnoses for these patients within 5 days of the date of the antidepressant prescription. We performed analyses for all eligible patients in PPRNet and for each practice. Results are reported using simple descriptive statistics. We used chi-square statistics for comparisons of nominal data.

Results

Demographic Information

Thirty-nine of the 68 PPRNet practices were eligible for the study. The 29 noneligible practices were excluded for the following reasons: enrollment in PPRNet after July 1, 1995 (21); specialty or urgent care practices (7); or did not reliably record diagnoses (1). Sixteen (41%) of the eligible practices were in rural environments; the remainder were in suburban or urban settings. The specialty distribution among the eligible practices included 32 family medicine (7 academic, 25 community), 6 general internal medicine, and 1 combined family medicine and internal medicine. A total of 389 physicians saw patients in the 39 practices during the study year.

There were 149,327 patients eligible for the study, 88,727 women (59.4%) and 60,272 men (40.4%). The gender was unknown for 328 (0.2%) patients. The age distribution was as follows: 54.7% aged 18 to 45 years; 26.0% aged 46 to 65 years; and 19.3 % aged 66 years or older. The median number of patients per practice was 2785 (range = 192 - 21,120). Socioeconomic information, such as occupation or insurance coverage, is not consistently maintained in the PPRNet database.

Diagnosis and Treatment of Depression

We excluded 18,186 (12.2%) of the 149,327 patients from this portion of the analysis. Among those excluded, 3231 (17.8%) had a previous diagnosis of depression, 7658 (42.1%) had previously received an antidepressant, and 7297 (40.1%) had been given a depression diagnosis and an antidepressant prescription before 1996. A greater number and proportion of women (13,329, 15.0%) than men (4852, 8.0%) were excluded.

A total of 2103 (1.6%) of the 131,141 patients without a documented history of depression or antidepressant prescription had a new diagnosis of depression in 1996 Table 1. There were 213 problem text strings in the electronic medical record that were bridged to depression for these 2103 patients. However, 72% were categorized as “depression,” and 6% were “depressive disorder.” Women were more likely than men to receive these diagnoses. Wide variation was found among the 39 practices in the incidence of depression (range = 0.4% - 4.0%).

Of the newly diagnosed patients, 1032 (49.1%) were started on antidepressants within 5 days of the initial visit for depression Table 1. Of those, 934 were given antidepressants alone, and 98 patients were also prescribed other psychotropic medications. Ninety-four newly diagnosed patients (4.5%) were prescribed only nonantidepressant psychotropic medication, and 977 (46.5%) patients not prescribed psychotropic medications within 5 days. Of the antidepressants prescribed within 5 days of depression diagnoses, 81% were SSRIs and 13% were TTAs. Of the other psychotropic medications prescribed, 73% were benzodiazepines and 14% were other antianxiety medications. Three hundred fifty-three (33.0%) of the 1071 patients diagnosed with depression who were not prescribed antidepressants within 5 days of their diagnoses received an antidepressant prescription by the end of 1996.

Among the 1032 patients given antidepressants, 69.9% had at least one definite follow-up appointment, and 90.5% had at least one least one definite or possible follow-up contact in the 6 months after their diagnosis. Among those with a follow-up, the mean number of contacts was 5.3.

Total Antidepressant Use

Overall, 9335 (6.3%) of the 149,327 patients received a prescription for an antidepressant in 1996 Table 2. Wide interpractice variability was present in the prescription of antidepressants (range = 1.9% - 13.6%). A total of 6443 diagnoses were made for these patients within 5 days of their antidepressant prescription. We present the distribution of these diagnoses in Table 3. One third of the patients (3045) who received antidepressants had a previous diagnosis of depression and had received antidepressants before 1996. Those patients were considered prevalent cases of depression and account for 2% of the total population. Four thousand twenty-two patients (more than 40% of the patients who received an antidepressant in 1996) never received a diagnosis of depression.

Discussion

The strength of our study was our ability to examine a relatively large number of primary care practices and patients, and to explore, in a longitudinal manner, an important mental health care issue in primary care. With appropriate caveats concerning the generalizability of findings from physicians who choose to use electronic medical records, this study adds additional information to the field’s growing body of knowledge.⁵

Our study confirmed that depression management and antidepressant prescribing are important aspects of primary care practice. During 1996, 1 of every 48 women and 1 of every 104 men who were patients in PPRNet practices received a new diagnosis of depression. Of those diagnosed with depression, almost half received an antidepressant prescription within 5 days of the diagnosis. By the end of 1996, nearly two thirds of these patients had received such a prescription, a proportion similar to that found in the standard care arm of a primary care depression trial.⁸ The vast majority of treated patients received SSRIs, a finding consistent with the increasing use of these agents among outpatient psychiatrists.⁹ SSRI antidepressants are better tolerated than older agents and do not increase overall treatment costs,¹⁰ justifying their widespread use.

Nine of every 10 patients treated with antidepressants had at least one follow-up contact with the practice. On average, these patients had more than 5 contacts in the 6 months immediately following the diagnosis. This degree of follow-up is consistent with the AHCPR guideline, which recommends contact every 10 to 14 days for the first 6 to 8 weeks of treatment and every 4 to 12 weeks after that.³ This finding is in contrast with the concern that this degree of follow-up is not provided in primary care settings.¹

Obscured in the overall analyses are the nearly 10-fold interpractice variations in the diagnosis of depression (0.8% to 8.6% for both 1996 and earlier diagnoses) and the use of antidepressants (1.9% to 13.6%). Patient-level factors may account for some of this variation. It is likely, though, that there are physician-level differences in the ability to recognize, the willingness to diagnose, and the comfort in the treatment of depression.^1,11 Excluding minor depression and dysthymia, which may be as prevalent, epidemiologic studies have found a prevalence of depression from 6% to 8% in primary care samples.¹² Only 4 of the 39 PPRNet practices in our study had a 6% or greater prevalence of depression. However, 21 practices prescribed antidepressants to 6% or more of their patients. Since many of the diagnoses for these patients are not accepted indications for antidepressants, it may be that physicians are treating more depression than they are recording. Nonetheless, it is clear that opportunities remain for improvement in depression recognition among nearly half of the study practices.

Limitations

Several limitations to our study are important. Diagnoses listed in the electronic medical record reflect physician opinion and are not validated against a gold standard. The PPRNet database has limited information about nonpharmacologic treatment, so the extent to which patients received counseling or watchful waiting is unknown. Both of these strategies are endorsed by the AHCPR guidelines for patients with mild depression.³ The nature of the database also makes it difficult to be certain about the duration of pharmacotherapy or patient compliance with prescribed medication. Finally, it is uncertain whether clinicians actually addressed the patient’s depression during follow-up contacts.

Conclusions

Our study reveals the interpractice variations of diagnosing depression and prescribing antidepressants in primary care. Patients with newly diagnosed depression were followed up in a manner consistent with published guidelines. Opportunities for increased detection and treatment of depression exist in approximately half of the study practices. The unique features of PPRNet provide opportunities for future work in this area.

Primary care clinicians, who by most estimates provide the majority of mental health care in the United States, are often criticized for the underdiagnosis and undertreatment of patients with depression.¹ However, most studies antedate the widely publicized 1993 Agency for Health Care Policy and Research (AHCPR) primary care practice guidelines.^2,3 A recent survey of 1350 primary care physicians documented diagnostic and treatment patterns consistent with high-quality care,⁴ but more information about actual practice is needed. Studies that examine depression management in the context of primary care using tools that enable the assessment of episodes of care are needed.⁵

The purpose of our study was to examine the initial pharmacologic management and follow-up of patients with newly diagnosed depression in primary care practice. The use of antidepressants among patients not diagnosed with depression was also examined. Our study was done within the Practice Partner Research Network (PPRNet), a national practice-based network of primary care physicians who use a common electronic medical record (Practice Partner, PMSI, Seattle, Wash) and pool data for research.⁶

Methods

We used a retrospective cohort design with the calendar year 1996 as the primary period of observation. Each month participating practices run a computer program to extract patient-level data from their electronic medical records. The first time the program runs, all existing data are extracted; in subsequent months only new data are extracted. To protect patient confidentiality, the program assigns a unique anonymous numerical identifier to each patient. Extracted data include diagnoses, medications, vital signs, laboratory results, and other variables; the text from progress notes and reports are not extracted. The information is copied to diskettes and mailed to the PPRNet coordinating center. At the center, data are aggregated and undergo rigorous quality control and bridging of text strings of diagnoses and medications to standard data dictionaries. Original text strings of diagnoses are maintained in the database. Updated data tapes are sent quarterly to the PPRNet offices at the Medical University of South Carolina Center for Health Care Research, where they are converted to SAS data sets (Statistical Analysis System, Cary, NC) on a UNIX computer. PPRNet staff use microcomputers with standard database and statistical software to perform all subsequent analyses.⁷

Practice eligibility criteria for our study included family medicine or general internal medicine specialty, enrollment in PPRNet before July 1, 1995, and reliable recording of diagnoses and medications. Patients were eligible for the study if they had at least one entry in their electronic medical records during the years 1993-1995 and were at least 18 years old by December 31, 1996.

We identified depression in patients with text strings on the problem list that bridged to depression diagnoses. Patients with one of these diagnoses or a prescription for an antidepressant medication dated earlier than January 1, 1996, were excluded from treatment analyses. We calculated incidence rates for depression as the number of new diagnoses divided by the number of eligible patients.

We classified psychopharmacologic agents as either antidepressants or other psychotropic medications. We subclassified antidepressants as (1) selective serotonin reuptake inhibitors (SSRIs); (2) tricyclic and tetracyclic agents (TTAs); or (3) other antidepressants. Other psychotropic medications were subclassified as (1) benzodiazepines; (2) other antianxiety agents; or (3) others, which included antipsychotic and antimania medications.

For new cases of depression, we defined a “definite follow-up” as another contact where a diagnosis of depression or another prescription for an antidepressant medication was recorded. Since physicians often address problems beyond those documented at each visit, we defined a “possible follow-up” as all definite depression follow-ups plus contacts for other diagnoses.

Initial analyses revealed that a large proportion of patients who received antidepressants in 1996 had never been diagnosed with depression. To explore this issue, we examined all diagnoses for these patients within 5 days of the date of the antidepressant prescription. We performed analyses for all eligible patients in PPRNet and for each practice. Results are reported using simple descriptive statistics. We used chi-square statistics for comparisons of nominal data.

Results

Demographic Information

Thirty-nine of the 68 PPRNet practices were eligible for the study. The 29 noneligible practices were excluded for the following reasons: enrollment in PPRNet after July 1, 1995 (21); specialty or urgent care practices (7); or did not reliably record diagnoses (1). Sixteen (41%) of the eligible practices were in rural environments; the remainder were in suburban or urban settings. The specialty distribution among the eligible practices included 32 family medicine (7 academic, 25 community), 6 general internal medicine, and 1 combined family medicine and internal medicine. A total of 389 physicians saw patients in the 39 practices during the study year.

There were 149,327 patients eligible for the study, 88,727 women (59.4%) and 60,272 men (40.4%). The gender was unknown for 328 (0.2%) patients. The age distribution was as follows: 54.7% aged 18 to 45 years; 26.0% aged 46 to 65 years; and 19.3 % aged 66 years or older. The median number of patients per practice was 2785 (range = 192 - 21,120). Socioeconomic information, such as occupation or insurance coverage, is not consistently maintained in the PPRNet database.

Diagnosis and Treatment of Depression

We excluded 18,186 (12.2%) of the 149,327 patients from this portion of the analysis. Among those excluded, 3231 (17.8%) had a previous diagnosis of depression, 7658 (42.1%) had previously received an antidepressant, and 7297 (40.1%) had been given a depression diagnosis and an antidepressant prescription before 1996. A greater number and proportion of women (13,329, 15.0%) than men (4852, 8.0%) were excluded.

A total of 2103 (1.6%) of the 131,141 patients without a documented history of depression or antidepressant prescription had a new diagnosis of depression in 1996 Table 1. There were 213 problem text strings in the electronic medical record that were bridged to depression for these 2103 patients. However, 72% were categorized as “depression,” and 6% were “depressive disorder.” Women were more likely than men to receive these diagnoses. Wide variation was found among the 39 practices in the incidence of depression (range = 0.4% - 4.0%).

Of the newly diagnosed patients, 1032 (49.1%) were started on antidepressants within 5 days of the initial visit for depression Table 1. Of those, 934 were given antidepressants alone, and 98 patients were also prescribed other psychotropic medications. Ninety-four newly diagnosed patients (4.5%) were prescribed only nonantidepressant psychotropic medication, and 977 (46.5%) patients not prescribed psychotropic medications within 5 days. Of the antidepressants prescribed within 5 days of depression diagnoses, 81% were SSRIs and 13% were TTAs. Of the other psychotropic medications prescribed, 73% were benzodiazepines and 14% were other antianxiety medications. Three hundred fifty-three (33.0%) of the 1071 patients diagnosed with depression who were not prescribed antidepressants within 5 days of their diagnoses received an antidepressant prescription by the end of 1996.

Among the 1032 patients given antidepressants, 69.9% had at least one definite follow-up appointment, and 90.5% had at least one least one definite or possible follow-up contact in the 6 months after their diagnosis. Among those with a follow-up, the mean number of contacts was 5.3.

Total Antidepressant Use

Overall, 9335 (6.3%) of the 149,327 patients received a prescription for an antidepressant in 1996 Table 2. Wide interpractice variability was present in the prescription of antidepressants (range = 1.9% - 13.6%). A total of 6443 diagnoses were made for these patients within 5 days of their antidepressant prescription. We present the distribution of these diagnoses in Table 3. One third of the patients (3045) who received antidepressants had a previous diagnosis of depression and had received antidepressants before 1996. Those patients were considered prevalent cases of depression and account for 2% of the total population. Four thousand twenty-two patients (more than 40% of the patients who received an antidepressant in 1996) never received a diagnosis of depression.

Discussion

The strength of our study was our ability to examine a relatively large number of primary care practices and patients, and to explore, in a longitudinal manner, an important mental health care issue in primary care. With appropriate caveats concerning the generalizability of findings from physicians who choose to use electronic medical records, this study adds additional information to the field’s growing body of knowledge.⁵

Our study confirmed that depression management and antidepressant prescribing are important aspects of primary care practice. During 1996, 1 of every 48 women and 1 of every 104 men who were patients in PPRNet practices received a new diagnosis of depression. Of those diagnosed with depression, almost half received an antidepressant prescription within 5 days of the diagnosis. By the end of 1996, nearly two thirds of these patients had received such a prescription, a proportion similar to that found in the standard care arm of a primary care depression trial.⁸ The vast majority of treated patients received SSRIs, a finding consistent with the increasing use of these agents among outpatient psychiatrists.⁹ SSRI antidepressants are better tolerated than older agents and do not increase overall treatment costs,¹⁰ justifying their widespread use.

Nine of every 10 patients treated with antidepressants had at least one follow-up contact with the practice. On average, these patients had more than 5 contacts in the 6 months immediately following the diagnosis. This degree of follow-up is consistent with the AHCPR guideline, which recommends contact every 10 to 14 days for the first 6 to 8 weeks of treatment and every 4 to 12 weeks after that.³ This finding is in contrast with the concern that this degree of follow-up is not provided in primary care settings.¹

Obscured in the overall analyses are the nearly 10-fold interpractice variations in the diagnosis of depression (0.8% to 8.6% for both 1996 and earlier diagnoses) and the use of antidepressants (1.9% to 13.6%). Patient-level factors may account for some of this variation. It is likely, though, that there are physician-level differences in the ability to recognize, the willingness to diagnose, and the comfort in the treatment of depression.^1,11 Excluding minor depression and dysthymia, which may be as prevalent, epidemiologic studies have found a prevalence of depression from 6% to 8% in primary care samples.¹² Only 4 of the 39 PPRNet practices in our study had a 6% or greater prevalence of depression. However, 21 practices prescribed antidepressants to 6% or more of their patients. Since many of the diagnoses for these patients are not accepted indications for antidepressants, it may be that physicians are treating more depression than they are recording. Nonetheless, it is clear that opportunities remain for improvement in depression recognition among nearly half of the study practices.

Limitations

Several limitations to our study are important. Diagnoses listed in the electronic medical record reflect physician opinion and are not validated against a gold standard. The PPRNet database has limited information about nonpharmacologic treatment, so the extent to which patients received counseling or watchful waiting is unknown. Both of these strategies are endorsed by the AHCPR guidelines for patients with mild depression.³ The nature of the database also makes it difficult to be certain about the duration of pharmacotherapy or patient compliance with prescribed medication. Finally, it is uncertain whether clinicians actually addressed the patient’s depression during follow-up contacts.

Conclusions

Our study reveals the interpractice variations of diagnosing depression and prescribing antidepressants in primary care. Patients with newly diagnosed depression were followed up in a manner consistent with published guidelines. Opportunities for increased detection and treatment of depression exist in approximately half of the study practices. The unique features of PPRNet provide opportunities for future work in this area.