Health Service Utilisation, Detection Rates by Family Practitioners, and Management of Patients with Common Mental Disorders in French Family Practice

Abstract

Objective:

Provide up-to-date detection rates for common mental disorders (CMD) and examine patient service-use since the Preferred Doctor scheme was introduced to France in 2005, with patients encouraged to register with and consult a family practitioner (FP) of their choice.

Methods:

Study of 1133 consecutive patients consulting 38 FPs in the Montpellier region, replicating a study performed before the scheme. Patients in the waiting room completed the self-report Patient Health Questionnaire (PHQ) and Client Service-Receipt Inventory with questions on registration with a Preferred Doctor and doctor-shopping. CMD was defined as reaching PHQ criteria for depression, somatoform, panic or anxiety disorder. For each patient, FPs completed a questionnaire capturing psychiatric caseness.

Results:

81.2% of patients were seeing their Preferred Doctor on the survey-day. Of those with a CMD, 52.6% were detected by the FP. This increased with CMD severity and comorbidity. Detected cases were more likely to be consulting their Preferred Doctor (84.7% versus 79.4% for non-detected cases, p = 0.05) rather than another FP. They declared more visits to psychiatrists (17.2% versus 6.7%, p = 0.002). There was no association with consultation frequency or doctor-shopping, which both declined between the two studies.

Conclusion:

The CMD detection rate is relatively high, with no increase compared to our previous study, despite a decline in doctor-shopping. An explanation is the same high proportion of patients visiting their usual FP on the survey-day at both periods, suggesting a limited impact of the scheme on care continuity. FP action taken highlights the importance of improving detection for providing care to patients with CMDs.

Common mental disorders (CMDs), most frequently defined as mood, anxiety, somatoform, and substance use disorders, remain a major public health problem, with 12-month prevalence rates ranging from 10% to 20% in high-income countries.¹ Most CMDs are managed in family practice,^2,3 where there is widespread evidence that approximately 50% of cases are not detected.^4–8 Although detection, mainly studied with regard to depression, has not always been shown to improve outcome,^5,9 it is a crucial and necessary first step in providing care whether through watchful waiting or immediate intervention.¹⁰

Detection of CMDs varies widely among family practitioners (FPs) within and between countries.^4,8,11 Detection is influenced by many factors, related to the FP,^12–14 the patient,^5,14–16 and the health care system.¹⁷ For example, detection appears to be higher in countries with more individualised care models characterised by a client-centred approach where there is continuity with the same FP.¹⁸ Detection increases with frequency of visits and when the patient is better known to the FP.^5,12 Paradoxically, detection has also been shown to be higher in doctor-shopping patients, but only among those consulting different FPs for reasons of dissatisfaction with previous care.¹⁹ Doctor-shopping is defined here as consulting a different FP to one’s usual FP; consulting a different FP is often unavoidable (e.g., FP away or patient on holiday) but in some cases is explained by the patient looking for a different approach.

Many detection studies^4,8,11 have been performed in countries with hierarchical primary health care models characterised by restricted access to care by different forms of gatekeeping, large group practices, and limited patient choice and access to a personal FP.^20,21 Little is known about France, where FPs work mainly alone on a fee-for-service basis.^22,23 In 2003, we carried out one of the only studies performed among randomly selected FPs using validated research instruments and found a 50% detection rate for CMDs.^6,19,24,25 At the time, patients could access any FP or specialist as often as they wanted, as illustrated by the high level of doctor-shopping (28.2% had seen at least 2 different FPs in the past 6 months, 18.4% for practical reasons and 9.8% for reasons of dissatisfaction).¹⁹

In 2005, a new scheme was introduced with the aim of structuring access to both primary and secondary care.²⁶ Patients (aged 16 years and above) register with an FP (or specialist doctor, although this is very rare: <1% of patients)²⁷ of their choice (the “Preferred Doctor”), who is responsible for coordinating care and referring the patient if necessary to secondary care (to all specialist doctors except gynaecologists, ophthalmologists, and psychiatrists for patients below the age of 25 years). To date, at least 95% of patients have registered with a Preferred Doctor,²⁶ and direct referrals have substantially declined.²⁸ Financial sanctions in the form of lower reimbursements (reduced from 70% to 30%) (www.ameli.fr) are applied to patients who consult: without having registered with a Preferred Doctor, with a different FP to their Preferred Doctor unless justified (FP unavailable, FP or patient away), or directly with a specialist doctor without being referred by the FP. These changes can be seen as a progressive move toward a more structured organisation of care in a system that still remains relatively open and patient-led. Indeed, patients can choose their doctor and can cancel and register with a new Preferred Doctor at will. However, the scheme encourages loyalty, which is based mainly on trust, quality of care, familiarity, and availability of the FP.²⁹ The scheme is likely to increase continuity of care between primary and secondary care and relational continuity with a single trusted doctor.^30,31 Patients are likely to be better known to their FP, and those with a CMD are more likely to be detected¹⁸ and accurately diagnosed.³² We thus hypothesize that FP detection of CMDs as well as continuity of care will have increased since the introduction of the Preferred Doctor scheme.

The aim of our study is to provide up-to-date prevalence and detection rates for CMDs, overall and by diagnostic category, within the new gatekeeping scheme. We also aim to describe the health and service-use characteristics of patients with CMDs, overall and according to whether they are detected by the survey day FP. Case management according to detection is also examined. Findings are interpreted in the light of those obtained in the 2003 study prior to the scheme.

Methods

FP Sample

The study was conducted from February to July 2013 in family practices in and around the city of Montpellier. We followed as closely as possible the same methodological approach as a previous study performed in 2003.⁶

In 2003, a random sample of 46 FPs drawn from the regional FP lists participated, with a response rate of 32.7% (see Norton et al⁶ for a detailed description of the recruitment method). For the 2013 study, all 46 FPs were contacted again; 34 were still practicing in the study area (6 had retired, 4 were not reachable, and 2 no longer worked in family practice). In all, 29 (85.3%) agreed to participate again. Those who refused stated a lack of time or interest; they were mainly males (4/5) practicing in the urban study zone (3/5). To adjust the age distribution of the sample to that of the FPs in the study area, 9 additional, younger FPs were recruited. We initially contacted the FPs who had taken over the practices of those who had retired or moved since 2003, and then we recruited 3 FPs from professional lists covering the study areas, approaching the youngest in age first. In all, this convenience sample included 38 FPs, of whom 29 had participated in the first study.

Patient Sample

In each FP practice, a research assistant invited all consecutive eligible patients entering the waiting room to participate. Patient inclusion criteria were: age (18 years and above) and the ability to read and write French. Patients and FPs signed a consent form prior to participation as required by the Montpellier University Hospital ethics committee. It was clearly explained that both patient and FP responses would remain confidential, with a common identification number linking patient and FP questionnaires.

Patients were recruited until approximately 30 per FP had participated. The response rate was 67%. The main reason for refusing was an objection to giving their name and address on the consent form.

None of the patients had participated in the 2003 study.

Instruments

Patient Data

Before the consultation, patients in the waiting room completed self-report questionnaires, including an adapted version of the Client Service Receipt Inventory (CSRI)³³ on service use and the Patient Health Questionnaire (PHQ).³⁴

The CSRI was adapted for the French care system, so information was collected on the number of FP consultations as well as the number of different FPs consulted over the past 6 months. Patients who had seen at least 2 different FPs (survey day visit included) were asked to give their reasons, classified as practical only versus dissatisfaction with previous care. Patients reported all hospitalisations, consultations with specialist doctors (type of specialist, number of visits), and other carers (nurses, physiotherapists, psychologists, etc.), along with any use of medication for psychological or sleeping problems. Questions were also added with respect to patients’ current status vis-a-vis the Preferred Doctor scheme.

The PHQ is a validated instrument that allows the identification of both threshold disorders (major depressive disorder, panic disorder, other anxiety disorder, bulimia nervosa) and subthreshold disorders (other depressive disorder, and the PHQ-specific categories: somatoform disorder, probable alcohol abuse/dependence, binge-eating disorders), by applying DSM-IV diagnostic criteria algorithms.^34,35 For somatoform disorders, patients meeting caseness criteria (applying the original algorithm to the somatoform module rather than PHQ-15 criteria)³⁵ and rated by the FP as moderately or severely physically ill were reclassified as noncases. This was the only way to ensure that those with a possible “adequate biological explanation” for the symptoms were not given the diagnosis. The PHQ had already been translated into French for the 2003 study.⁶ Patients were considered as having a CMD if they reached criteria on the PHQ for a somatoform disorder, major or other depressive disorder, panic disorder, or other anxiety disorder.

Patients also completed a form on their sociodemographic characteristics, and level of disability was measured using the Brief Disability Questionnaire.⁸

FP Data

During the consultation, the FPs completed a short questionnaire adapted from the World Health Organisation Physician’s Encounter Form.³⁶ The FPs were asked to estimate the severity of physical and psychiatric illness, regardless of presenting symptoms, by ticking a box on a 5-point scale ranging from completely healthy to severely ill, with a clear indication that ratings of 3 and above (mild, moderate, or severe) indicated caseness and required a diagnosis (open script). The FPs then indicated what actions were undertaken, such as providing listening and support, from a predefined list.

Statistical Analysis

Frequencies and medians (minimum-maximum) were used to describe the FP and patient samples. FPs who had participated in the previous study (n = 29) were compared with those recruited for the current study only (n = 9) by using Wilcoxon’s rank sum test and the chi-square test. For patients, all comparative analyses were performed using marginal Generalised Estimating Equations (GENMOD SAS procedure, option repeated) to take into account the 2-stage sampling process and clustering of patients by practice. Statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, North Carolina).

Results

The FP sample is described in Table 1. As expected, the 9 “new” FPs were younger (p = 0.0003) and, except for 2, had all practiced for less than 10 years. They were more likely to be in part-time practice (66.7% versus 17.2% for those who participated in both studies, p = 0.009), reflecting the higher proportion of female FPs (77.8% versus 34.5%, p = 0.05).

Table 1.

Description of family practitioner (FP) sample.

	2013 study (N = 38)
	n	%
Practice setting
Urban	27	71.1
Semi-rural	11	28.9
Sex
Male	21	55.3
Female	17	44.7
Age (median [minimum-maximum])	38	51.5 [31-65]
FP since
<10 years	7	18.4
≥10 years	31	81.6
Group practice
No—solo	14	36.8
Yes—partner FP(s)	24	63.2
Type of practice
Full-time	27	71.1
Part-time	11	28.9
Mental health training (past 3 years)
No	18	47.4
Yes, evening only	5	13.2
Yes, daytime seminar	15	39.5

Among the patients, 35% were male, and 43% were aged 50 or above. Half (51%) were married or living with someone, and 31% were single. On the survey day, 81.2% of patients were seeing their Preferred Doctor and 94.8% reported having a Preferred Doctor, of which 44.9% had already changed Preferred Doctor at least once since 2005. Among the reasons for changing, 15.4% reported dissatisfaction with a previous FP and 3.3% lack of availability of FP; other reasons were patient or FP had moved (59.7%), FP had retired (16.5%), and distance to FP practice (5.1%).

Of the patients, 25.5% met PHQ criteria for a CMD: 10% for a somatoform disorder; 7.2% and 9.6% for other and major depression, respectively; 5.2% for panic disorder; and 6.9% for other anxiety disorder. The FPs classified 30.6% of the patients as having a psychiatric disorder. The main diagnoses given were anxiety and/or panic disorder (43%), depression (21%), and anxiodepressive symptoms (17%). The FPs detected 52.6% of CMD cases (95% CI, 46.8 to 58.4), increasing with severity, from 44.2% for somatoform disorders to 79.7% for panic disorder (Figure 1). The detection rate for major and other depression combined was 60.3%; for a score of 10 or higher on the depression module (PHQ-9), the detection rate was 57.4%. The detection rate increased with the number of comorbid disorders: 46.6% for 1 disorder only, 53.7% for 2 disorders, and 82% for 3 or 4 disorders (collapsing major and minor depression and excluding alcohol and eating disorders) (p < 0.0001). Of the patients not reaching CMD criteria, 77.1% (95% CI, 74.2 to 79.9) were classified as noncases by the FP (specificity). Case detection was stratified according to whether the FPs had participated in the 2003 study: Among the 29 initial FPs, 47.9% (95% CI, 41.2 to 54.6) of cases were detected, for 66.4% (95% CI, 55.6 to 77.2) among the 9 “new” FPs (p = 0.007).

Figure 1.

Proportion of patients per Patient Health Questionnaire (PHQ) diagnostic category: overall (light + dark grey) and detected by the family practitioner (FP) (dark grey). Percentages correspond to FP detection rates per category. Bolded categories are included in the study definition of common mental disorder.

Patients with a CMD seeing their Preferred Doctor on the survey day were slightly more likely to be detected (p = 0.05) (Table 2). Similarly, they were less likely to be detected if seen by the FP for the first time (p = 0.02) (Table 3).

Table 2.

Patients’ sociodemographic characteristics and service use, according to CMD status and FP detection.^a

						CMD = yes (n = 287)
	CMD = no (n = 839)		CMD = yes (n = 287)			CMD not detected (n = 136)		CMD detected (n = 151)
	n	%	n	%	p	n	%	n	%	p
On the survey day
Sex: male	320	38.1	75	26.1	<0.0001	36	26.5	39	25.8	0.72
Age
<30 y	191	22.8	71	24.7		44	32.4	27	17.9
30-49 y	265	31.6	112	39.0		50	36.8	62	41.1
50-64 y	185	22.1	65	22.7		21	15.4	44	29.1
>65 y	198	23.6	39	13.6	<0.0001	21	15.4	18	11.9	0.007
Educational level
Low (<9 y)	170	20.3	62	21.7		47	34.6	75	49.7
Medium (9-12 y)	220	26.2	80	28.0		64	47.1	61	40.4
High (>12 y)	449	53.5	144	50.4	0.54	25	18.4	15	9.9	0.82
Marital status
Single	244	29.1	104	36.2		57	41.9	47	31.1
Marital life	453	54.0	121	42.2		58	42.7	63	41.7
Divorced/separated	90	10.7	45	15.7		12	8.8	33	21.9
Widowed	52	6.2	17	5.9	0.003	9	6.6	8	5.3	0.03
Living situation
Aloneb	273	32.5	122	42.5		47	34.6	75	49.7
With partnerb	466	55.5	125	43.6		64	47.1	61	40.4
Other	100	11.9	40	13.9	0.002	25	18.4	15	9.9	0.004
Occupation
Working	402	48.0	123	42.9		59	43.4	64	42.4
Unemployed	79	9.4	43	15.0		20	14.7	23	15.2
Retired	223	26.6	49	17.1		22	16.2	27	17.9
Other	133	15.9	72	25.1	<0.0001	35	25.7	37	24.5	0.96
Disability level (score out of 11)
Low (0-1)	330	39.5	38	13.2		23	16.9	15	9.9
Moderate (2-5)	298	35.7	85	29.6		42	30.9	43	28.5
High (>5)	208	24.9	164	57.1	<0.0001	71	52.2	93	61.6	0.25
Personal and social problemsc
No	333	39.8	41	14.3		19	14.0	22	14.6
A little	382	45.7	99	34.5		52	38.2	47	31.1
A lot	121	14.5	147	51.2	<0.0001	65	47.8	82	54.3	0.22
Psychological reason for FP visitd	75	8.9	78	27.2	<0.0001	16	11.8	62	41.1	<0.0001
Seeing preferred doctor
Yes	677	80.8	235	82.2		108	79.4	127	84.7
No	116	13.8	39	13.6		19	14.0	20	13.3
Does not have one	45	5.4	12	4.2	0.71	9	6.6	3	2.0	0.05
In the past 6 months (CSRI)
Hospitalisation: yes	113	13.5	57	19.9	0.006	23	17.0	34	22.5	0.07
Visit to specialist doctor: yes	426	50.9	188	65.7	<0.0001	80	59.3	108	71.5	0.06
Visit to psychiatrist: yes	35	4.2	35	12.2	<0.0001	9	6.7	26	17.2	0.002
Visit to MHP: yes	90	10.7	70	24.5	<0.0001	23	17.0	47	31.1	0.006
Doctor-shopping in FP
No	705	84.3	235	81.9		108	79.4	127	84.1
Yes, practical reason	101	12.1	35	12.2		19	14.0	16	10.6
Yes, dissatisfied reason	30	3.6	17	5.9	0.12	9	6.6	8	5.3	0.48
No. of visits to an FP ≥7	61	7.3	41	14.3	0.005	19	14.0	22	14.6	0.57
Medicatione: yes	181	21.6	140	48.8	<0.0001	43	31.6	97	64.2	<0.0001

^aLess than 2% of missing values for all variables. CMD = common mental disorder; CSRI = Client Service Receipt Inventory; FP = family practitioner; MHP = mental health professional (psychiatrist, psychologist, or psychotherapist).

^bWith or without children.

^cBothered by.

^dVisiting FP for a psychological reason or for a problem with a psychological cause.

^eFor psychological reason or sleeping problems.

Table 3.

FP management of patients, according to CMD status and FP detection.^a

						CMD = yes (n = 287)
	CMD = no (n = 839)		CMD = yes (n = 287)			CMD not detected (n = 136)		CMD detected (n = 151)
	n	%	n	%	p	n	%	n	%	p
Patient seen for first time	81	9.7	27	9.5	0.98	19	14.1	8	5.4	0.02
FP action at the end of the visit
Prescription of anxiolytic or antidepressant medication	51	6.1	46	16.1	<0.0001	2	1.5	44	29.1	<0.0001
Prescription of any medication	617	73.5	227	79.1	0.08	110	80.9	117	77.5	0.41
Further investigation of somatic complaint	128	15.3	41	14.3	0.53	20	14.8	21	13.9	0.99
Referral to a specialist doctor	109	13.0	57	19.9	0.002	20	14.8	37	24.5	0.18
Referral to a psychiatrist	3	0.4	7	2.4	0.008	0	0.0	7	4.6	NV
Referral to another health care professional	51	6.1	12	4.2	0.30	5	3.7	7	4.6	0.69
Appointment made for follow-up visit	78	9.3	34	11.9	0.13	15	11.1	19	12.7	0.27
Discussion of problems, listening and support	202	24.1	97	33.9	0.0006	31	23.0	66	43.7	<0.0001
Therapeutic education	75	8.9	37	13.0	0.10	14	10.5	23	15.2	0.25
Other (medical certificate)	124	14.8	36	12.6	0.49	16	11.9	20	13.3	0.83
Duration of visit
<10 min	39	4.7	11	3.8		9	6.6	2	1.3
10 min	245	29.4	59	20.6		31	22.8	28	18.5
10-20 min	453	54.4	166	57.8		85	62.5	81	53.6
20-30 min	63	7.6	38	13.2		6	4.4	32	21.2
>30 min	33	4.0	13	4.5	0.034	5	3.7	8	5.3	0.0002

^aCMD = common mental disorder; FP = family practitioner; NV = test not valid.

Discussion

Our study confirms the high prevalence of CMDs seen in family practice.^8,23,34,37 Detection rates, overall and for specific diagnoses, appear to be higher than those found elsewhere,^4,34,38,39 when considering that our definition of CMDs includes subthreshold disorders. The detection rate is in keeping with our previous study,²³ not confirming our hypothesis of an increase linked to the introduction of the Preferred Doctor scheme.⁴⁰ While doctor-shopping has declined in the 10-year gap, the proportion of the patients seeing their usual or Preferred Doctor on the survey day has not changed. Patients with a CMD are more likely to be detected when they report higher secondary care service use. This is not the case for those who reported high attendance rates and doctor-shopping in family practice.

The detection rate is similar to that found in our previous study (51%; 95% CI, 45.5 to 56.6).⁴⁰ When the sample is stratified according to participation in the previous study, the rate for the 29 initial FPs is very close to that obtained 10 years earlier. This could suggest stability over time in FP strategies and thresholds for detecting CMDs. The higher rate among the “new” FPs could be explained by differences in age, sex, and years of practice between the 2 subsamples, although there is little evidence in our study or elsewhere of a link between age and sex of FP and detection. It could also reflect greater emphasis on mental health during FP training.⁴¹ However, caution is required in interpreting our findings given the low number of cases per FP available for detection. We believe this accounts for the wide confidence intervals, rather than the wide variability in detection rates between FPs reported elsewhere.^4,8,11

The lack of increase in detection rates between the 2 studies could be explained by the relatively high level of detection already found in 2003 (50%), especially considering that cross-sectional study designs tend to underestimate case detection.^11,42 A cautious approach to detection is common; FPs preferring watchful waiting⁴³ and discussion of symptoms to increase patient awareness and acceptation⁴⁴ rather than immediate diagnosis and intervention. This is particularly true for mild conditions, as shown in our study where detection rates increase with the severity of disorders^14,45 and comorbidity.⁴⁶ Also, patients may chose not to disclose symptoms, and a normalising attribution style hinders and postpones detection.⁴⁷ Last, studies often use categorical diagnostic classification systems, such as DSM, as their reference for detection. This approach is not always considered useful to FPs who prefer a dimensional approach based on symptom severity.⁴⁸

In our 2013 study, we reported higher detection rates than in the PHQ validation study for mood (60.3% versus 51%) and anxiety disorder (66.7% versus 43%).³⁴ Detection rates for major depression are also higher than in a recent large-scale study also using the PHQ (57.4% versus 51% for a PHQ-9 score ≥10).⁴ A meta-analysis on mild depression reported a detection rate of 33.8% compared with 46% in our study.³⁸ For anxiety, our rates of 80% for panic disorder and 72% for other anxiety disorder can be compared to an overall rate obtained by meta-analysis of 44.5%.³⁹

As reported elsewhere, our study shows CMDs to be linked to female gender, middle-age groups, marital status, occupation, disability, and personal and social problems.^8,14,49,50 These characteristics are mirrored by higher FP detection rates for the 30- to 49-year age group,^15,51 being divorced or separated, and living alone. Patients visiting for psychological reasons are more likely both to have a CMD and to be detected by the FP, which suggests that openly presenting psychological symptoms facilitates detection and increases its likelihood.^47,52

Regarding service use, patients hospitalised or consulting a specialist doctor or psychiatrist in the past 6 months were more likely to meet CMD criteria.² This is mirrored by a higher detection rate, despite being borderline significant for hospitalisation and number of specialist doctor visits.

Both number of FP visits²⁵ and doctor-shopping¹⁹ have declined considerably since the 2003 study.⁴⁰ In 2013, doctor-shopping was associated neither with having a CMD nor with FP detection; conversely, our findings from the 2003 study showed a higher proportion of CMD patients among doctor-shoppers whether for practical or dissatisfaction reasons. FP detection was also higher for doctor-shoppers but only for those reporting dissatisfaction with previous care.¹⁹ CMD patients consulted FPs more frequently, in keeping with the literature^53,54 and our previous study.²⁵ Frequent visits often reflect unmet needs⁵⁵ or a higher level of patient needs.⁵⁴ However, contrary to findings from elsewhere,¹² this was not linked to detection in our study. This may be explained by a difference in the definition of frequent attendance, in terms of recall period, number of visits, and the lack of age and sex stratification of attendance.⁵⁶

We found that seeing one’s Preferred Doctor on the survey day was associated with a slightly higher detection rate, in keeping with studies suggesting that FPs are better able to detect CMDs in patients with whom they are more familiar.^12,16 Similarly, detection was lower for patients seen by the FP for the first time. Doctor-shopping had declined since the previous study, which could suggest a greater continuity of care. However, in each study, approximately the same proportion of patients declared seeing their usual or Preferred Doctor on the survey day,²⁵ meaning that the scheme may only have formalised what already existed,⁵⁷ with a limited impact on relational care continuity. This contradictory finding could be tentatively explained by the already high proportion of patients consulting their usual FP on the survey day prior to the scheme and possibly by different participation rates between patients seeing or not seeing their usual FP. There could also be a looser interpretation of what was considered as a “usual” FP in 2003 compared with the Preferred Doctor in 2013, which requires registration. Yet the lack of increase in continuity of care since the introduction of the scheme, as measured by the status (usual doctor or not) of the survey day FP, may explain our failure to find a significant increase in detection between the 2 surveys.

The most frequent action undertaken by the FP on the survey day was a prescription of medication. Patients reaching CMD criteria were more likely to be referred to a specialist doctor, specifically to a psychiatrist; there was no association with FP detection, which may be due to small numbers. CMD status and detection were both associated with longer consultation times, confirming that patients with psychiatric symptoms require more time.⁵⁸ This is difficult to interpret in a cross-sectional study design as it is not possible to know whether detection increased due to longer visits or the opposite. In so far as longer consultations can be seen as beneficial to the patient,⁵⁹ our findings highlight the importance of detection for providing adequate care.

The main limitation of our study is the lack of a gold standard for case-detection. Indeed, CMD caseness is based on self-reported symptoms, despite using the DSM-IV criteria-based PHQ designed specifically to overcome the difficulties of obtaining interview-based diagnoses in primary care. Also, the sample cannot be seen as representative of FPs practicing in the study area because of the initial low response rate (32.7%), although it is comparable to that reported in much family practice research.⁶⁰ FPs particularly interested in mental health may have selectively agreed to participate, perhaps indicated by the retention rate for the 2003 FPs (85.3%). Cross-sectional designs for studying detection can be criticised because they tend to underestimate detection.¹¹ Finally, comparisons between the 2003 and 2013 studies require caution given the other possible explanations for differences beyond the Preferred Doctor scheme.

The main strength of our study is the relatively large number of patients with a high participation rate, for whom data are available from both the patient and the FP. Unfortunately, this was not sufficient to study inter-FP detection rates given the limited number of CMD cases per FP. We used validated research tools, making our findings comparable to those of other studies. Also, the presence of a research assistant in the waiting room yielded high-quality data. We provide up-to-date findings on detection with the originality of comparable data from a previous study carried out using a similar approach in an overlapping sample of FPs.⁴⁰ This is one of the only studies to provide comparable data obtained before and after the introduction of change in access and use of health care services.

The high but expected prevalence of CMDs in our study is coupled with a relatively high FP detection rate, similar to that in our previous study. Although doctor-shopping has declined since the introduction of the Preferred Doctor scheme, we saw no change over the 10-year period in the already high proportion of patients seeing their usual FP on the survey day. This suggests a limited impact of the scheme on relational continuity of care. It could explain the lack of increase in detection following the implementation of the scheme in 2005. Detected cases benefitted from longer consultation times with more support and advice and were more likely to receive prescriptions for psychotropic medication. This highlights the importance of improving detection as a crucial first step in tackling the growing public health and economic burden of CMDs.⁶¹ Further research is needed to better understand what interventions could help improve detection and patient management.