Management of chronic headache with referral from primary care to direct access to MRI compared with Neurology services: an observational prospective study in London.

OBJECTIVES: To evaluate the cost, accessibility and patient satisfaction implications of two clinical pathways used in the management of chronic headache. INTERVENTION: Management of chronic headache following referral from Primary Care that differed in the first appointment, either a Neurology appointment or an MRI brain scan. DESIGN AND
SETTING: A pragmatic, non-randomised, prospective, single-centre study at a Central Hospital in London. PARTICIPANTS: Adult patients with chronic headache referred from primary to secondary care. PRIMARY AND SECONDARY OUTCOME MEASURES: Participants' use of healthcare services and costs were estimated using primary and secondary care databases and questionnaires quarterly up to 12 months postrecruitment. Cost analyses were compared using generalised linear models. Secondary outcomes assessed: access to care, patient satisfaction, headache burden and self-perceived quality of life using headache-specific (Migraine Disability Assessment Scale and Headache Impact Test) and a generic questionnaire (5-level EQ-5D).
RESULTS: Mean (SD) cost up to 6 months postrecruitment per participant was £578 (£420) for the Neurology group (n=128) and £245 (£172) for the MRI group (n=95), leading to an estimated mean cost difference of £333 (95% CI £253 to £413, p<0.001). The mean cost difference at 12 months increased to £518 (95% CI £401 to £637, p<0.001). When adjusted for baseline and follow-up imbalances between groups, this remained statistically significant. The utilisation of brain MRI improved access to care compared with the Neurology group (p<0.001). Participants in the Neurology group reported higher levels of satisfaction associated with the pathway and led to greater change in care management.
CONCLUSION: Direct referral to brain MRI from Primary Care led to cost-savings and quicker access to care but lower satisfaction levels when compared with referral to Neurology services. Further research into the use of brain MRI for a subset of patient population more likely to be reassured by a negative brain scan should be considered. TRIAL REGISTRATION NUMBER: NCT02753933. © Author(s) (or their employer(s)) 2020. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.

The estimate of healthcare resource use was based on comprehensive and complete data retrieved from hospital databases supplemented by both primary care utilisation data and self-reported participant data.

The evaluation of the intervention’s impact considered different dimensions of analysis (efficiency, quality of care, access to care and patient satisfaction).

This was a single-centre study with participants recruited from one central hospital in London, therefore a multicentre study would be necessary to explore the generalisability of the results.

Introduction

Globally, the percentage of the adult population with an active headache disorder is 47% for episodic headache and 3% for chronic headache (lasting more than 15 days per month).1 Headache is in the top 10 international causes of disability,2 with nearly half of sufferers reporting it affects work, home or social activities.3–5 Most headaches are primary headache disorders, such as migraine or tension-type headaches. Secondary headaches, due to an underlying serious pathology (eg, tumour, brain aneurysm) are far less common.6 In fact, less than 0.1% of the lifetime headache prevalence is associated with a life-threatening condition, which can include primary or secondary brain tumours.2 3

Most headache sufferers self-manage, but over 4% of adults each year consult their general practitioner (GP).7 8 GPs manage 97% of headache presentations, particularly GPs with special interest and training in headache training initiatives,9 with 2% of these referred to neurologists and 1% to other specialists.7 Headache is the most common cause for GP referrals to neurologists accounting for up to 20%–30%,10–13 the vast majority of these are for migraine. Chronic migraine sufferers (>15 days/month) had more emergency department (ED)/hospital visits, and diagnostic tests than those with episodic migraine and consequently, the medical costs were three times higher.14 Hence, despite the benign nature of most headaches, headache management is associated with high healthcare utilisation. Furthermore, a study in multiple countries of the European Union found that headache management was variable in terms of visits with GPs and specialists and medication use.15 Between 2012 and 2014, patients presenting to headache specialists (either neurologists or GPs with a special interest in headache) costed £956 million in the UK.13 16

GP direct access to imaging has been defined as a priority within the National Health Service (NHS), with direct access to brain MRI for the diagnosis of brain cancer identified as a specific initiative.8 GPs have reported referrals for secondary care, both for a neurologist consultation or neuroimaging, when they were unable to reassure the patient.17 18 Furthermore, Morgan et al found that referral for headache is often the outcome of patient pressure and anxiety interacting with GP characteristics, organisational factors and service availability rather than the headache severity itself.17 This contrasts with the recommendations from the National Institute for Health and Care Excellence that does not recommend the use of neuroimaging for reassurance purposes.19 A US study estimated that patients with new onset migraine headache or a flare-up of chronic headache had, respectively, a 39% (95% CI 24% to 54%) and 51% (95% CI 32% to 68%) probability of having neuroimaging routinely ordered even where guidelines specifically recommended against this approach.20 A UK-based randomised controlled trial evaluated the cost implications of using brain MRI for reassurance purposes and found that it led to a decrease in healthcare costs, in particular for patients with higher levels of psychiatric morbidity.21

In summary, despite proportionately low level of referrals to secondary care, high prevalence makes headache the most frequently listed reason for referral to neurologists and thus, uses capacity that is severely constrained. The potential decrease of referral rates to Neurology specialists as a result of direct access to imaging holds the potential release resources. On the other hand, direct access may be associated with a transfer of workload to Radiology, as not all patients directly referred to brain MRI would routinely undergo an MRI scan as part of the management of their chronic headache. This study aims to evaluate the healthcare utilisation and associated costs of GPs using direct access to brain MRI for the management of chronic headache patients compared with referral to a Neurologist.

Patients and methods

Study design and patient selection

The study was an independent single-centre site, prospective, observational study. The study compared two existing clinical pathways in the management of patients with a chronic headache from either GP referral to the Neurology Department or to direct access to brain MRI. No change to clinical practice was involved. Participants were allocated to two groups, the Neurology group or the MRI group. The allocation was decided a priori, that is, the referrer (in this case the GP) decided which referral route would suit each participant. Subsequent care was consistent with a standard of care for each clinical pathway and depicted in figure 1.

Figure 1

High-level illustration of two existing clinical pathways associated with the referral from GP due to chronic headache. GP, general practitioner.

Patients eligible for the study included adults aged 16 years or over with chronic migraine defined as occurring ≥15 days per month for more than 3 months (as per International Classification of Headache Disorders 3 (ICHD-3) criteria) and referred from GP practices to a local Hospital in Central London, either for a Neurology outpatient appointment or an MRI examination. Participants were excluded if there were secondary causes for the headache, if they were prisoners, lacked the capacity to give consent or participate in the study, not fluent in English or already taking part in a clinical trial of an investigational medicinal product. Following eligibility assessment, participants completed a written informed consent.

Patient and public involvement statement

This study has involved patient and public involvement during the design and dissemination phase. Prior to the start of the research, a group of patients suffering from chronic headache were consulted to coproduce the study mobile app that was used to monitor symptoms and triggers of chronic headache. The patients who were involved in the study design, as well as all study participants, received a lay summary of the research findings as per their preferences, either via post or email. Research data are available on reasonable request.

Study outcomes

The primary outcome was to estimate the 6-month healthcare costs associated with two existing clinical pathways in the management of chronic headaches. Secondary outcomes comprised: the extension of the cost analysis up to 12 months, the evaluation of access to care, patient satisfaction, headache burden and time off work associated with both clinical pathways.

Service use costs

Total costs were calculated based on the multiplication of any headache-related healthcare events by the unit cost of each event. Resource use data included contacts with any NHS healthcare provider associated with the management of chronic headache. These included, among others, visits to GPs or headache clinical nurse specialist, inpatient care, neurologist or other headache-related outpatient visits (eg, psychiatry), physiotherapist, visits to the ED, advanced imaging such as CT and MRI. Resource use data were retrieved from multiple hospital and primary care databases as well as self-reported data from participants using a participant headache diary (either paper-based or a headache mobile app). For the purposes of the primary outcome, the valuation of unit costs was, whenever possible, based on NHS Reference Costs 2016-17.22 Medication costs were derived from Prescription Cost Analysis23 and estimated from clinical data, specifically secondary care clinic letters and information provided by primary care. Participants were also asked to record time off work due to headache using the weekly participant diary.

Quality of life and headache burden

Headache burden was assessed using validated headache questionnaires and headache diaries. The HIT-6 questionnaire measured the headache burden based on six questions, leading to a score range from 36 to 78. The Migraine Disability Assessment Scale (MIDAS) questionnaire assessed the disability associated with the headache (ranges from 0 to 90). Furthermore, along with the headache diaries completed by participants, the MIDAS questionnaire allowed the estimate of the number of headache days per quarter (maximum of 90 days) and self-reported headache pain scores (ranging from 0, no pain at all, to 10, the worst pain ever).

Access to care

The time elapsed between the GP referral and the first appointment in each group, either the Neurology appointment or the MRI scan, was measured as a proxy for access to care. Furthermore, in the MRI group, the availability of the MRI report to the GP, rather than the actual timing of the diagnostic scan, was considered.

Change in care management

Change in care management in both groups was evaluated. It was considered that a change in care management occurred when patients underwent new treatment options (medication or otherwise, eg, new headache medication, Botox treatment).

Patient satisfaction

Patient satisfaction was evaluated at month 3 postrecruitment using a non-validated questionnaire. This questionnaire evaluated three dimensions of analysis: (a) referral process (time elapsed between referral from primary care to initial secondary care appointment); (b) initial appointment and (c) overall experience 3 months after recruitment.

Statistical analyses

This study was observational but all analyses were based on the principle of ‘intention-to-treat’ so that participants recruited were included in the analysis as per the group they were recruited to, regardless of whether they actually received the intended treatment, any protocol deviations or potential losses to follow-up.24 Given the study’s time horizon of 12 months, no discounting of costs or effects were considered. Descriptive statistics on several socio-demographics baseline characteristics were included. Differences in baseline characteristics between treatment groups were compared using statistical tests: χ2 for categorical data, t-test for Normal data and Mann-Whitney for non-Normal quantitative data. Given the skewness associated with the cost distribution, all cost differences between groups were assessed using generalised linear models (GLM) with an identity-link and gamma distribution. An identity link function instead of a log link was considered in order to avoid potential analytical biases.25 26 An unadjusted GLM cost analysis with the study group (MRI group vs Neurology group) as only univariate analysis was performed as the first step. Given the study’s observational design, the cost difference can be impacted due to the lack of randomisation.27 Baseline variables that differenced between the two groups (p<0.1) were included in the multivariable (adjusted) GLM analyses. For all GLM analyses, group difference estimates and associated confidence intervals were reported, together with p values. Analyses were conducted using Stata V.15.

Sample size

The sample size estimate was calculated based on the primary endpoint, total 6-month healthcare costs. A total of 150 participants were recruited in the Neurology group and 99 participants in the MRI group to achieve a detection a cost difference of £300 assuming SD of £750 and £500, respectively, with 85% power at the 5% two-sided significance level. A 20% increase in sample size due to unknown cost distribution and attrition rate was considered.

Results

A total of 249 participants were recruited, 150 in the Neurology group and 99 in the MRI group. In total, 100% (n=128) and 97% (n=96) of participants recruited received the treatment allocation in the Neurology and the MRI group, respectively (figure 2). With regards to the follow-up duration, 15% (n=22) and 4.0% (n=4) participants withdrew from the Neurology and MRI group, respectively, and were considered lost to follow-up. Remaining participants (n=223) were included in the analysis, equivalent to 128 (85%) and 95 (96%) participants in the Neurology and MRI group, respectively.

Figure 2

Participant flow chart for the headache study.

Baseline sociodemographic, clinical variables and resource use in the 12 months prerecruitment are detailed in table 1. A higher proportion of women were recruited to the Neurology group compared with the MRI group (81% vs 68%, p=0.039). Both groups were similar in mean age, ethnic mix, with more than half being White. No significant differences between the groups were found in terms of highest qualification and employment, with high-school and full-time employment being the most common. In terms of clinical variables, no significant difference between groups were found in terms of number of active health problems or number of headache triggers. Participants in the Neurology group reported lower utility and health scores using the generic 5-level EQ-5D (EQ-5D-5L) questionnaire, with a statistically significant difference in health scores (p=0.005). Lower health scores imply lower self-rated quality of life, while for the headache-specific questionnaires (HIT-6 and MIDAS), a higher score implies a higher headache burden. Compared with the MRI group, participants in the Neurology group reported significantly higher headache burden (HIT-6 mean score: 65 vs 63, p=0.006), number of headache days in the past 3 months (MIDAS questionnaire: 52 vs 43, p=0.038). Participants in the Neurology group also had significantly more primary care appointments (3.7 vs 2.4, p<0.001) and all NHS appointments (4.3 vs 2.5, p<0.001).

Table 1

Baseline characteristics of the population analysed

Variables	Neurology group (n=128)	MRI group (n=95)	P value
Age, years, mean (SD)	38.4 (14.1)	40.0 (14.6)	0.514
Gender, Female, n (%)	103 (81)	65 (68)	0.039
Ethnicity			0.079
Asian	13 (10)	6 (6.3)
Black	40 (31)	17 (18)
White	68 (53)	60 (63)
Mixed	5 (3.9)	8 (8.4)
Other	2 (1.6)	4 (4.2)
Qualification, n (%)			0.565
Advanced work or PhD	3 (2.3)	4 (4.2)
Master’s degree	15 (12)	12 (13)
Bachelor’s degree	31 (24)	29 (31)
High school	57 (45)	33 (35)
Did not finish high school	8 (6.3)	10 (11)
Prefer not to answer	14 (11)	7 (7.4)
Employment, n (%)			0.839
Employee in full-time job (30 hours or more a week)	50 (39)	36 (38)
Employee in part-time job (under 30 hours a week)	18 (14)	18 (19)
Self-employed, full or part time	12 (9.4)	8 (8.4)
Full-time education at school, college or university	16 (13)	9 (9.5)
Doing something else	2 (1.6)	2 (2.1)
Permanently sick/disabled	8 (6.3)	6 (6.3)
Looking after the home	4 (3.1)	6 (6.3)
Unemployed and available for work	12 (9.4)	6 (6.3)
Wholly retired from work	4 (3.1)	4 (4.2)
Prefer not to answer	2 (1.6)	0 (0)
Presence of mental health condition, n (%)	22 (17)	11 (12)	0.243
Number of active health problems: mean (SD)	2.0 (1.5)	1.8 (1.4)	0.277
Number of headache triggers: mean (SD)	2.1 (1.8)	1.8 (1.4)	0.378
Self-reported questionnaires
EQ-5D-5L, mean utility (SD)	0.809 (0.182)	0.830 (0.195)	0.097
EQ-5D-5L, mean score (SD)	64 (19)	71 (20)	0.005
HIT-6, mean score (SD)	65 (5.3)	63 (7.3)	0.006
MIDAS, mean score (SD)	58 (54)	45 (45)	0.075
MIDAS, mean headache days (SD)	52 (32)	43 (31)	0.038
MIDAS, mean pain score (SD)	6.9 (1.8)	6.9 (1.9)	0.778
Resource use in the 12 months prior to recruitment
GP appointments, mean (SD)	3.7 (2.9)	2.4 (1.5)	<0.001
All NHS events, mean (SD)	4.3 (3.7)	2.5 (1.5)	<0.001

EQ-5D-5L, 5-level EQ-5D; GP, general practitioner; HIT-6, Headache Impact Test; MIDAS, Migraine Disability Assessment Scale; NHS, National Health Service.

Service use

Table 2 summarises the NHS resource use of primary care and hospital-based services over a period of 12 months following recruitment. With regards to primary care utilisation, participants in the Neurology group had a significantly higher number of GP face-to-face appointments (mean number per participant: 1.82 vs 1.19, p=0.006). Regarding secondary care, participants in the Neurology group had a higher mean number of outpatient appointments (2.52 vs 0.26, p<0.001) and other treatments such as Botox and nerve root injection (mean 0.30 vs 0.05, p<0.001). In contrast, they had a lower mean number of brain MRIs (0.59 vs 1.05, p<0.001). There were no statistically significant differences in the utilisation levels of any of the other healthcare events.

Table 2

Breakdown of number of NHS appointments per type of activity organised per group and respective number of participants responsible for these appointments (12 months postrecruitment)

Type of NHS appointment	Neurology group (n=128)			MRI group (n=95)			P value
	Total of episodes	Mean (SD)	N (%)	Total of episodes	Mean (SD)	N (%)
Primary care services
GP face-to-face appointment	233	1.82 (2.11)	91 (71)	113	1.19 (1.64)	57 (60)	0.006
GP phone appointment	37	0.29 (0.75)	24 (19)	25	0.26 (0.49)	23 (24)	0.420
Hospital-based services
Hospital outpatient appointment	322	2.52 (1.19)	128 (100)	25	0.26 (0.55)	20 (21)	<0.001
Inpatient episode	4	0.03 (0.35)	1 (0.8)	1	0.01 (0.10)	1 (1.1)	0.837
Emergency department episode	8	0.06 (0.24)	8 (6.3)	5	0.05 (0.22)	5 (5.3)	0.756
Head CT	1	0.01 (0.09)	1 (0.8)	1	0.01 (0.10)	1 (1.1)	0.832
Brain MRI	75	0.59 (0.49)	75 (59)	100	1.05 (0.30)	95 (100)	<0.001
Others (eg, botox and nerve injection treatments)	39	0.30 (0.79)	25 (20)	5	0.05 (0.30)	3 (3.2)	<0.001

GP, general practitioner; NHS, National Health Service.

The NHS resource use pre and postrecruitment for both groups was also compared. Table 3 summarises the difference between headache-related events 12 months postrecruitment compared with the 12 months prerecruitment. A reduction of 197 and 87 GP visits (mean 1.54/0.92 per participant or a percentage reduction of 85%/77%) was noted in the Neurology and MRI group, respectively. Similarly, there was a reduction in ED utilisation with a decrease of 23 and 6 episodes (mean 0.18 and 0.07 per participant or a percentage reduction of 288% and 120%) in the Neurology and MRI group, respectively.

Table 3

Difference in the number, mean and percentage reduction of NHS appointments per type of activity organised per group 12 months postrecruitment compared with the 12 months prerecruitment (note: a negative/positive percentage denotes a decrease/increase in activity following recruitment)

Type of NHS appointment	Neurology group (n=128)			MRI group (n=95)
	Total of episodes	Mean	%	Total of episodes	Mean	%
Primary care services
GP face-to-face appointment	−197	−1.54	−85	−87	−0.92	−77
GP phone appointment	−11	−0.09	−30	5	0.05	20
Hospital-based services
Hospital outpatient appointment	301	2.36	93	23	0.24	92
Inpatient episode	3	0.02	75	1	0.01	100
Emergency department episode	−23	−0.18	−288	−6	−0.07	−120
Head CT	−9	−0.07	−12	1	0.01	1.0
Brain MRI	55	0.43	141	100	1.05	NA

GP, general practitioner; NHS, National Health Service.

Cost analyses

The mean (SD) cost management per participant at 6 months postrecruitment was lower in the MRI group compared with the Neurology group (£245 (£172) vs £578 (£420)), leading to a mean cost difference between groups of −£333 per participant (95% CI: −£413 to −£253, p<0.001). The MRI group had a higher proportion of participants with lower costs (£0–£250 range) than the Neurology group (73% vs 4.7%). The initial cost analysis was extended to 12 months postrecruitment and the mean cost difference between groups increased, with the MRI group generating cost savings of £518 per participant (95% CI: −£637 to −£401, p<0.001).

Given the non-randomised study design, an additional GLM analysis was performed to adjust for baseline characteristics. The unadjusted 6-month cost difference between groups (−£333; 95% CI: −£413 to −£253) hardly changed after adjustment (−£308; 95% CI: −£408 to −£209). In all analyses, at both 6 and 12 months postrecruitment, direct access to MRI for the management of chronic headache was associated with statistically significant mean cost savings for the NHS.

Abnormal findings in the MRI group

Out of the 95 participants recruited in the MRI group, three MRIs were not performed during the initial appointment due to claustrophobia events. From a total number of 92 MRIs, 85 (92%) were normal and 7 (7.6%) presented abnormal findings (table 4), with one diagnosis being particularly significant (two small intracranial aneurysms). This participant was referred to neurovascular team for assessment, at which point no intervention was performed during the follow-up period (participant included in an active surveillance group). Two other participants had a change in their clinical management for less significant findings, leading to either a follow-up appointment or scan. No brain malignancies were diagnosed.

Table 4

Description of incidental findings, its clinical relevance and subsequent pathway

Abnormal findings	Significant (yes/no)?	Changes in diagnostic or treatment pathway
Mature striatocapsular lacune	No
Sinusitis with complete opacification	No	Ear, nose, and throat (ENT) specialist review only
Pituitary abnormality (T1)	No
Low lying cerebellar tonsils	No
Previous petrous surgery noted	No
Two aneurysms AcomA (anterior communicating artery) and right ICA	Yes	Referred to neurovascular—no coiling (no intervention)
No definitive lesion	No	Follow-up MRI only

ICA, internal carotid artery.

The mean time (SD) elapsed from GP referral to the initial Neurology and MRI scan was, respectively, 110 (35) days and 39 days (17), a statistically significant difference (p<0.001). Second, if the MRI report, than the actual scan, is considered as a proxy for access to care, the mean time elapsed was 70 days (35) (p<0.001).

Change in management

Participants in the Neurology group were more likely to have a change in therapeutic management compared with participants in the MRI group (97% vs 64%, p<0.001). Similarly, among participants that had not started on preventative medication prerecruitment, a higher proportion of participants in the Neurology group were started on preventative medication as part of their clinical management (93% vs 53%, p<0.001).

At baseline, mean utility was lower in the Neurology group but not significant (mean utility of 0.809 vs 0.830, p=0.097), while their self-perceived mean health score was significantly lower (EQ-5D-5L: visual analogue scale score of 64.0 vs 70.8, p=0.005) (table 5). There was no statistically significant differences between the groups at 6 months for the utility and self-reported healthcare scores.

Table 5

Descriptive statistics for one generic questionnaire (EQ-5D-5L) and two headache-specific questionnaires (HIT-6 and MIDAS) at baseline and 6 months postrecruitment

				N	Mean	SD	P value
Baseline	EQ-5D-5L	Utility	Neurology	127	0.809	0.182	0.097
			Radiology	95	0.830	0.195
		Score	Neurology	126	64.0	18.8	0.005
			Radiology	95	70.8	20.2
	HIT-6	Score	Neurology	128	65.0	5.3	0.006
			Radiology	92	62.6	7.3
	MIDAS	Score	Neurology	124	57.8	54.0	0.075
			Radiology	90	44.8	44.9
		Headache days	Neurology	124	51.6	31.5	0.038
			Radiology	90	42.8	30.7
		Pain score	Neurology	124	6.9	1.8	0.778
			Radiology	90	6.9	1.9
Month 6	EQ-5D-5L	Utility	Neurology	55	0.770	0.263	0.243
			Radiology	26	0.681	0.346
		Score	Neurology	53	68.2	20.6	0.463
			Radiology	23	62.5	24.7
	HIT-6	Score	Neurology	35	60.0	8.7	0.968
			Radiology	12	53.1	22.8
	MIDAS	Score	Neurology	32	52.6	58.1	0.827
			Radiology	12	40.7	36.0
		Headache days	Neurology	46	41.2	28.6	0.152
			Radiology	15	26.5	28.1
		Pain score	Neurology	44	4.5	2.3	0.663
			Radiology	15	4.2	2.0

EQ-5D-5L, 5-level EQ-5D; HIT-6, Headache Impact Test; MIDAS, Migraine Disability Assessment Scale.

At baseline, participants in the Neurology group presented a higher headache burden compared with the MRI group (mean score of 65.0 vs 62.6, p=0.006) (table 5). This trend seemed to be maintained over the follow-up period but was difficult to interpret due to high attrition rates: mean HIT-6 score at 6 months (60.0 vs 53.1, p=0.968). At baseline, participants in the Neurology group had higher headache severity compared with the MRI group, but this was not significant (mean MIDAS score of 57.8 vs 44.8, p=0.075) (table 5). In both groups, the headache severity decreased over the follow-up period. At baseline, participants in the Neurology group reported a significantly higher mean number of headache days compared with the MRI group (51.6 vs 42.8, p=0.038). This trend was observed at month 6 but was not significant (41.2 vs 26.5, p=0.152). A third variable assessed self-reported headache pain scores. At baseline, participants at baseline showed almost identical mean headache pain scores (6.9 vs 6.9, p=0.827). During the follow-up period, no statistical differences were found at month 6 (4.5 vs 4.2, p=0.663).

Time off work

Participants in the Neurology group (n=83) had a higher mean number of days off work due to headache compared with participants in the MRI group (n=35) but these were not statistically significant at 6 (13.9 vs 9.7, p=0.563) or 12 months (27.9 vs 19.1, p=0.808) postrecruitment.

Patient satisfaction was evaluated at month 3 postrecruitment using a non-validated questionnaire (online supplemental file 1). Participants in both groups reported no difference in terms of receiving an appointment in a suitable timeframe (p=0.193). Participants in the MRI group reported higher satisfaction levels (p=0.005) compared with the Neurology group associated with the information received prior to the actual appointment. No statistically significant difference (p=0.366) between the groups was found regarding the satisfaction levels of both appointments (either MRI scan or Neurologist appointment) but a higher proportion of participants in the Neurology group reported a better experience compared with their expectation (p=0.002). At month 3 postrecruitment in all variables, except frequency of appointments (p=0.166), participants in the Neurology group reported higher levels of satisfaction with: amount of time spent with clinical staff (p=0.001); consistency of care (p=0.028); how informed you felt about your condition (p=0.010)/your treatment (p=0.004) and the overall experience (p<0.001).

Discussion

Previous UK studies assessed the use of direct access to advanced imaging in the management of chronic headache patients.10 21 Howard et al21 performed a randomised controlled trial, which showed that the use of imaging in patients led to a reduction of referral rates to Neurology services in secondary care from 23% to 1.3% (1/76) in the treatment group (ie, patients being scanned).21 Thomas et al10 estimated that direct access to brain imaging, in this case CT, reduced referral rates to Neurology in 86% of the cases during the follow-up period (average of 1.3 years per patient).10

The study’s underlying hypothesis was that the early use of an advanced and accurate diagnostic tool (in this case MRI) would reassure both patients and GPs that no serious underlying cause (particularly brain tumour) is present. This would, in turn, reduce the headache burden and NHS resource use associated with the patient’s subsequent management. Given the high prevalence of headache and the increased referral of patients with chronic headaches and other neurological conditions from primary care to hospital-based care,7 it is relevant to assess the implications of using different management strategies. To our knowledge, no previous prospective study in the UK has assessed the economic implications of these two coexisting management strategies based on GP referral decision.

The primary outcome was total costs at 6 months postrecruitment. The study showed that the use of advanced imaging produced cost-savings to the NHS compared with referral to Neurology, with mean cost-savings per participant of £333 and £518 at month 6 and 12 postrecruitment, respectively (p<0.001). These cost differences were multifactorial but primarily driven by: (1) the lower unit cost of a brain MRI scan (£146) compared with the initial Neurology appointment (£240); (2) the lower number of outpatient appointments in the MRI group (25 vs 322); (3) the fact that 75/128 (59%) of participants in the Neurology group ended up having a brain MRI scan in the 12-month period of follow-up and (4) the increased likelihood of patients in the Neurology arm receiving management with Botox injections or nerve block procedures (39 vs 5 events). Direct access to brain MRI seemed to reassure most participants as only 17 (18%) participants in the MRI group ended up being referred to a neurologist. Furthermore, over 66% of participants in the MRI group had no further hospital-based care, compared with 5% in the Neurology group. Similarly, at primary care level, participants in the MRI group presented lower utilisation rates per participant when compared with the Neurology group (mean GP appointments of 1.82 vs 1.19, p=0.006). In order to mitigate the potential confounding by indication introduced by GPs referring patients to both clinical pathways, adjusted GLMs were conducted. All cost differences between groups remained statistically significant (p<0.001) when adjusted for differences in baseline characteristics or follow-up attrition rates.

At baseline, participants in the Neurology group presented lower self-reported quality of life and higher headache burden. HIT-6 scores did improve over time but we were unable to assess whether there was a statistically significant difference in both groups due to the high attrition rates, particularly in the MRI group. As mentioned, out of the 95 participants recruited to the MRI group, 17 (18%) were subsequently referred to a Neurologist. Interestingly, at baseline, these 17 patients reported higher headache burden compared with all 95 patients recruited to the MRI group (MIDAS score: 51.2 vs 44.8; MIDAS headache days: 55.8 vs 42.8). This finding seems to suggest that these data might be useful to risk stratify patients and support GPs in their referral criteria. However, further research into this area is required.

Participants in the Neurology group presented a higher probability of being started on a preventative medication and had a change in therapeutic management following referral from primary care. A small proportion (7.6%) of participants in the MRI group had abnormal findings in the initial brain MRI. Only one participant had clinically significant lesions (two small aneurysms) which were also incidental and no brain tumour was diagnosed.

Time off work was also evaluated as a proxy of headache burden. Participants in the Neurology group presented a trend (p>0.05) of a higher number of days off work due to headache compared with participants in the MRI group. This difference was not, however, statistically significant.

Lastly, patient satisfaction in both groups was compared based on three dimensions of analysis: during the referral period, the initial appointment and overall satisfaction. Twenty-five per cent of participants in the Neurology group (vs 18% in the MRI group) reported dissatisfaction with the waiting time. This finding is not unexpected as the mean waiting time associated with the Neurology appointment was almost three times the one associated with the MRI scan (110 vs 39 days). Contrary to the referral period, participants in the Neurology group reported trends of higher satisfaction levels associated with the first appointment (neurology outpatient visit vs MRI scan) and a better experience compared with their expectations (60% vs 29% in the Neurology and MRI group, respectively). Participants in the Neurology group reported improved satisfaction levels at 3 months across different variables (time spent with clinician, consistency of care, information about the condition and its treatment). Almost three quarters of Neurology participants reported being satisfied or very satisfied with their headache management compared with only 21% in the MRI group (p<0.001). Both participants and GP referrers in the MRI group reported dissatisfaction associated with the waiting time elapsed between the MRI scan and the availability of results (mean time of 31 days). This might have contributed to increased anxiety in some participants.

Strengths of this study

The estimate of NHS resource use data was primarily based on comprehensive and complete data retrieved from hospital-based databases that captured both the acute and elective elements of the pathway associated with the management of patients with chronic headache. These data were supplemented by both primary care utilisation data, collected from each participant’s GP and self-reported participant data. The aim was to guarantee that any chronic headache-related NHS event was costed regardless of the healthcare provider and its location. The prospective collection of healthcare utilisation and the evaluation of the impact of the interventions across different dimensions of analysis (efficiency, quality of care, access to care and patient satisfaction) were other key factors that contributed to the overall strength of the study.

Limitations of this study

There were some limitations to this study. First, this was a single-centre study with participants recruited from one central hospital in London. A multicentre study would be necessary to explore the generalisability of the results. Second, as with any observational study, no randomisation between groups was performed and there were significant differences in baseline of headache burden and health-scores and utilisation of resources. In order to mitigate potential confounding factors, adjusted multivariate analyses showed that the primary outcome was hardly affected and remained statistically significant. Nevertheless, residual confounding factors remain a limitation of the study. Third, specific inclusion and exclusion criteria were considered and as such, the study sample might not be representative of all patients with chronic headache. Fourth, for the purpose of secondary outcomes, most data were self-reported and hence prone to recall bias. Lastly, there were high follow-up attrition rates, particularly for participants in the MRI group, which affected some of the study’s secondary outcomes.

Conclusions

This study found that the referral from primary care to direct access to brain MRI compared with referral for a neurologist for patients with chronic headache was associated with lower NHS overall costs at 6 and 12 months postrecruitment. Despite waiting longer from referral to appointment, participants in the Neurology group reported higher satisfaction levels associated with the care received compared with the MRI group and were more likely to benefit from changes to their therapeutic management.

Implications for further research and clinical practice

Baseline measures of headache burden, such as the HIT-6 or MIDAS, could potentially be used to determine which pathway may be suitable for patients but further research into the risk stratification of chronic headache patients is required. Future study designs should consider the confounding by indication introduced by recruiting participants referred to the two pathways based on GP decision. With regards to clinical practice, and based on the study findings, we plan to provide GPs with headache management advice along with the MRI report.