Comprehensive Inpatient Treatment of Refractory Chronic Daily Headache


  • Mathematical symbols: χ2 = chi square; F = F value (analysis of variance); t = Student's t test (independent samples); P = probability.

  • Conflict of Interest: None

A.E. Lake III, Michigan Head-Pain and Neurological Institute, 3120 Professional Drive, Ann Arbor, MI 48104, USA.


Objective.— (1) To assess outcome at discharge for a consecutive series of admissions to a comprehensive, multidisciplinary inpatient headache unit; (2) To identify outcome predictors.

Background.— An evidence-based assessment (2004) concluded that many refractory headache patients appear to benefit from inpatient treatment, underscoring the need for more research, including outcome predictors.

Methods.— The authors completed a retrospective chart review of 283 consecutive admissions over 6 months. The inpatient program (mean length of stay = 13.0 days) included intravenous and oral medication protocols, drug withdrawal when indicated, cognitive-behavior therapy, and other services when needed, including anesthesiological intervention. Patient-reported pain levels and consensus of medical staff determined outcome status.

Results.— The 267 completers (94%) included 212 women and 55 men (mean age = 40.3 years, range = 13-74) from 43 states and Canada. The modal diagnosis was intractable, chronic daily headache (85%), predominantly migraine. Most (59%) had medication overuse headache (MOH), involving opioids (48%), triptans (16%), or butalbital-containing analgesics (10%). Psychiatric diagnoses included stress-related headache (82%), mood disorders (70%), anxiety disorders (49%), and personality disorders (PD, 26%). More patients with a PD (62%) had opioid-related MOH than those with no PD (38%), P < .005. Of the completers, 78% had moderate to significant pain reduction, with comparable improvement in mood, function, and behavior. Clinical factors predicting moderate-significant headache improvement were limited to MOH (84% vs 69%, P < .007) and presence of a PD (68% vs 81%, P < .03).

Conclusions.— Most patients (78%) improved following aggressive, comprehensive inpatient treatment. Maintenance of improvement is likely to depend on multiple post-discharge factors, including continuity of care, compliance, and home or work environment.

There exists a group of headache sufferers, many of whom become dependent on opioids, whose pain and impairment remain refractory to outpatient treatment. In most cases of chronic headache, daily opioids fail to provide sustained relief or reduce functional impairment.1,2 Frequent opioid use is likely to contribute to central sensitization and progressive worsening of the headache through multiple mechanisms.3,4

Meaningful headache control, improved functioning and enhanced quality of life require a level of care intensity that matches the severity of illness.5 For patients with medication overuse headache (MOH) involving opioids, butalbital-containing analgesics, triptans, or ergotamine tartrate, successful drug withdrawal with aggressive headache treatment may require inpatient level care. An evidence-based assessment (2004) concluded that many intractable headache patients benefit from inpatient treatment, and underscored the need for more research, including identification of outcome predictors.6

We previously published 6- to 8-month outcome studies of patients admitted to our inpatient unit,7 and later as part of a comprehensive evaluation of both inpatient and outpatient outcome for our entire center.8 The population referred has become increasingly complex over time, often with multiple psychiatric and medical comorbidities.

This study addresses 2 objectives:

  • 1To assess outcome at discharge for a large consecutive series of admissions from a national referral base, and
  • 2To identify clinical factors predicting successful outcome.


The overall headache care system involved in this study includes both outpatient and inpatient programs for refractory head, neck, and facial pain. Patients were first evaluated at the outpatient center. Only patients who met explicit criteria were admitted to the inpatient unit: daily or near daily headache at a level of severity that significantly compromised function and activities of daily living, failure of aggressive outpatient therapy, and in the case of MOH, failed efforts at drug withdrawal as an outpatient. For some patients, there were also potential safety issues associated with treatment or drug withdrawal that required close monitoring. Inpatient services include individually tailored intravenous protocols; oral prophylactic and abortive medication; withdrawal from overused medications; cognitive-behavioral therapy with family intervention by licensed PhD psychologists, relaxation/distraction therapy; and psycho-educational groups. When indicated, patients also receive interventional anesthesiology, physical and occupational therapy, and consultation for comorbid conditions. The milieu includes 24-hour staffing by trained nurses exclusively assigned to the unit, with daily multidisciplinary team meetings.

Following approval by a national Institutional Review Board with concurrence of appropriate personnel at Chelsea Community Hospital (which does not have an internal review board for scientific studies), one of the authors (A.E.L.) reviewed medical record data for 283 consecutive patients treated on the unit over a 6-month period. Of this group, 267 (94%) completed the program, 9 requested early but orderly discharge, 5 exited the program against medical advice, and 2 transferred to inpatient psychiatry. All patients participated in the basic program components; 134 also received anesthesiological intervention. The mean length of stay (LOS) for the year under study was 13.0 days.

Outcome analyses for program completers were based on data from the structured discharge summary. Outcome ratings of improvement in pain, function, and behavior (no change, minimal, moderate, or significant) were based on patient global pain ratings 4 times/day, and consensus of clinical staff working with the patient, including the primary attending physician (program director – J.R.S.) and assistant director (R.L.H.). If rating differences occurred between staff and patient, observations of patient functioning swayed the direction of the final rating. The first author (A.E.L.) performed all statistical analyses with Systat 8.0 (Systat Software, Inc., San Jose, CA, USA).

The final discharge headache diagnoses were determined by the program director (J.R.S.), based on the International Classification of Headache Disorders, 2nd Edition (ICHD-II)9 and the new ICHD-II appendix criteria open for a broader concept of chronic migraine (CM), including MOH.10 Psychiatric diagnoses were determined by licensed clinical psychologists, based on criteria from the Diagnostic and Statistical Manual of the American Psychiatric Association, 4th Edition, Text Revision (DSM-IV-TR).11 The final discharge psychiatric diagnoses relied on a combination of interview data from a comprehensive psychological evaluation at Michigan Head-Pain and Neurological Institute prior to admission, further clinical interviews and treatment by the inpatient program psychologist who worked with the patient during his or her stay, and observation of patient behavior on the unit.


Demographics.— Completers (N = 267) came from a total of 43 states and Canada, including 46.4% from the tri-state area of Michigan (90), followed by Ohio (19) and Indiana (15). The sample included 212 women and 55 men, with a mean age of 40.3 years (median = 41, SD = 13.4, range = 13-74). Of this group, 105 were in the work force (90 employed, 16 self-employed); 79 were not working by choice or at least partially due to headache, 38 were on disability, 18 retired, and 27 were students. Payors included Blue Cross of Michigan (42, 15.7%), Blue Cross of other states (72, 27.0%), commercial insurance (108, 40.5%), Medicare (36, 13.5%), Federal insurance (5, 1.9%), and private pay (4, 1.5%).

Headache Diagnoses and Medical Comorbidities.— The modal primary diagnosis was intractable chronic daily headache (CDH), typically CM (226). The other primary diagnoses included chronic posttraumatic headache (11), new daily persistent headache (9), cervicogenic headache (6), cluster headache (6), facial pain (5), hemicrania continua (3), and occipital neuralgia (1). Most (158) also met criteria for MOH with prescription drugs, primarily opioids (127), followed by triptans (39) and ergots (4), and butalbital compounds (28); 61 overused multiple medications. Two or more significant comorbid medical conditions occurred in 240, including 95 with comorbid non-headache painful disorders.

Psychiatric Diagnoses.— All patients had Axis I clinical syndromes, primarily stress-related physiological response affecting a medical condition (219), mood disorders (188), and anxiety disorders (132); 115 had both a mood and anxiety disorder. Axis II personality disorders (PDs) were identified in 70, primarily Cluster B (borderline, histrionic, narcissistic, antisocial = 42), followed by Cluster C (avoidant, dependent, obsessive-compulsive = 31); 17 patients met diagnostic criteria from both Clusters B and C; 13 had a PD not otherwise specified. Cluster A disorders (paranoid, schizoid, schizotypal) were rare, present in only 2 patients who also met criteria for either a Cluster B or Cluster C disorder.

Outcome.—Table 1 depicts the level of improvement at discharge in pain, mood, function, and behavior, for all headache diagnoses combined. For CDH-CM, 76.9% attained moderate or clinically significant pain improvement, compared with 82.9% of patients with other primary headache diagnoses [χ2(1) = 0.73, n.s.]. As noted above, 50% of the patients required at least one neural blockade, based on appropriate clinical criteria.

Table 1.—. Level of Improvement for All Headache Diagnoses Combined (Discharge Summary Data)
DomainRated level of improvement
Pain112 (42.1%)95 (35.7%)30 (11.3%)29 (10.9%)
Mood103 (38.7%)101 (38.0%)31 (11.7%)31 (11.7%)
Function109 (41.0%)93 (35.0%)31 (11.7%)33 (12.4%)
Behavior102 (38.3%)98 (36.8%)30 (11.3%)36 (13.5%)

The treatment team found effective abortive/rescue medication for 74.1% of the patients, and identified additional behavioral treatment goals for 87.6%. For the 59 patients who experienced minimal or no overall pain improvement at discharge, the team identified behavioral treatment goals for 96.6%, other treatment options and referrals for 89.8%, and effective abortive/rescue medication for 23.7%.

MOH: Drug Withdrawal and Outcome.— All program completers were successfully withdrawn from overused medication. Of the 158 patients with MOH, 84% attained moderate to significant improvement in headache control, compared with 69% of the 109 non-overusers [χ2(1) = 7.39, P < .007]. The percentage of patients with moderate to significant pain improvement at discharge was greatest for triptan-dependent MOH (38/39 = 97.4%) and ergot dependence (4/4 = 100%), followed by opioid dependence (103/127 = 81.1%) and butalbital-containing analgesics (22/28 = 78.6%). Combining the categories of triptan and ergot overuse, the differences in improvement rates for the 3 types of MOH were significant [χ2(2) = 7.50, P < .025].

MOH and LOS.— Although patients with any prescription drug overuse (ie, with single or multiple categories of prescription abortives) had better outcomes, they also had a significantly longer LOS than those with no prescription drug MOH on admission, a mean difference of 1.8 days [t(258) = 2.63, P < .009]. This was primarily due to a mean LOS that was 2.7 days longer for opioid-dependent patients vs those with no opioid dependence [t(241) = 3.99, P < .00009].

An analysis of variance of LOS for 4 groups of patients – opioid dependence with a comorbid PD, opioid dependence with no PD, PD only, and those with neither opioid dependence nor a PD – showed significant differences [F(3) = 8.33, P < .00003]. Post hoc analysis comparing each of the groups using Bonferroni adjustment revealed that patients with both a PD and opioid overuse had a mean LOS that was significantly longer than each of the other 3 conditions, by a mean of 2.9 days vs patients with opioid dependence only (P < .03), 5.2 days vs patients with a PD but no opioid dependence (P < .0015), and 4.6 days longer than those with neither opioid dependence nor a PD (P < .00003). In contrast, post hoc analysis did not show the LOS for the other 3 groups to differ from each other.

Although it might appear intuitive that LOS would be determined by the payor, for the completers as a whole this was not the case. An analysis of variance of LOS for the 4 primary payors (in-state Blue Cross, out-state Blue Cross, Medicare, and commercial insurance) was not statistically significant [F(3) = 0.33, P < .80]. Although financial constraints may have reduced LOS for some individual patients, any physician recommendations for continued hospital days were solely based on clinical need.

Clinical Factors Predictive of Outcome: Personality Disorders.— Pain improvement at discharge was unrelated to demographics, the number of comorbid medical conditions, presence of other painful comorbidities, or presence of a mood or anxiety disorder. However, comparisons involving the presence or absence of a PD diagnosis found that patients with PDs were more often opioid-dependent (62.3% vs 37.7%, χ2(1) = 8.12, P < .0045), and less likely to reach at least moderate improvement by discharge [68.1% vs 81.2%, χ2(1) = 5.08, P < .025]. As shown in Table 2, the presence of a PD, with and without MOH, was a significant negative prognostic indicator [χ2(3) = 15.49, P < .0015].

Table 2.—. Clinical Factors Predictive of Outcome: Axis II Personality Disorders and Prescription Medication Overuse
MOHAxis II PDn%Improvement at dischargeχ2P
Moderate to significantMinimal or none
  1. PD = personality disorder; MOH = prescription medication overuse headache.

NonePD176.49 (52.9%)8 (47.1%)15.49<.0015
NoneNone9134.266 (72.5%)25 (27.5%)
MOHPD5239.938 (73.1%)14 (26.9%)
MOHNone10619.594 (88.7%)12 (11.3%)


Roughly 4 out of 5 (78%) of a national sample of patients with complex, severe, and treatment-resistant cases of CDH, who completed this comprehensive multidisciplinary inpatient program, experienced at least moderate improvement in overall pain by the time of discharge. Comparable improvement occurred for mood, function, and behavior. Outcome was not statistically related to the number of comorbid medical conditions, the presence of other non-headache painful conditions, or demographic variables, although the presence of other clinical variables affected treatment complexity and the estimated costs of care.

Interestingly, patients with MOH were significantly more likely to achieve clinical efficacy than those with no analgesic or abortive medication dependence. The best outcomes occurred in triptan-dependent patients, where 97% reached at least moderate improvement. Overuse of opioids, analgesics, butalbital-containing combinations, and triptans/ergots confounds treatment of the primary headache disorder, and drug withdrawal is generally believed necessary to achieve the best headache control, including recovery of therapeutic responsiveness.12 The improvement in MOH patients is particularly encouraging, if not counterintuitive, given data suggesting that significant resolution of MOH may require several weeks to months of abstinence from the withdrawn drug, particularly in those overusing opioids.12-14

Other than the presence of MOH, the only patient variable predicting outcome was the presence or absence of a PD. The percentage of patients with a PD in this sample of nationally referred intractable headache patients study is much higher than what has been reported in US epidemiological studies of the general population, which identify about 9-11% of the population meeting criteria for at least one PD.15-17 The 26% prevalence in this study is actually quite comparable to the 24% prevalence of PDs in a consecutive sample of primary care attendees in the United Kingdom18 and somewhat less than the 35% PD prevalence in a series of patients with either somatoform pain disorder or other medical illnesses with long-standing work disability at a rehabilitation clinic.19

The lower rate of improvement at discharge in the PD group may be due to a number of factors. Cluster B (borderline, histrionic, narcissistic, and antisocial) was the most common group of PDs in our study, diagnosed in 60% of the patients with any type of PD. Patients with borderline PD experience dysregulation of both affective and nociceptive systems. Several studies show attenuated perception of externally induced experimental pain in subjects with borderline PD compared with those without borderline features.20-22 A pain patient with borderline PD may be far more attentive to internally induced pain and distress, with significant deficits in his or her ability to self-regulate or tolerate distress without escalation.23,24 Both the clinical experience of seasoned headache specialists25 and recent research data26 identify headache sufferers with borderline PD as more difficult to manage from a doctor–patient relationship perspective, with more treatment-resistant headaches.

The significantly higher rate of opioid dependence in the PD group compared with the non-PD group is particularly curious and may be due to a variety of interesting factors. Physicians may be more likely to administer opioids to demanding, insistent patients than to less aggressive patients with equal clinical severity.27,28 Certain patients are likely to be intense in their demands and expectations of the physicians. In the case of patients with a PD from Cluster B, aggressive requests for opioids may be based on dysregulated pain-related affect, emotional distress in general, and the need to control relationships, as well as narcissistic entitlement beliefs. These behaviors, more than the severity of headache itself, may prompt the physician to administer opioids defensively.

The neurobiology of borderline behavior may involve mu receptor mediation,29 which might explain the relationship we found in this study between presence of a PD, opioid-related MOH, and outcome. Borderline PD is significantly associated with substance disorders in both adults30 and adolescents.31 Opioids may not only induce MOH in the Cluster B group but may also aggravate other adverse psychological symptoms, including flashbacks, dissociation, the subjective sense of “emptiness,” and self-injurious behavior. Opioid receptor blockade has been shown to ameliorate dissociation phenomena in borderline patients, supporting a neurobiological linkage between opioids and certain features of borderline psychopathology.32

Successful drug withdrawal in many patients with PD may only be accomplished in an inpatient setting where control and boundaries can be established. For example, in an evaluation of outcomes from a highly structured outpatient program for a consecutive series of headache patients with psychiatrically diagnosed borderline PD, the percentage with MOH remained unchanged between baseline (74.0%) and after 6 months of treatment (76.7%). In contrast, for a matched control group of migraine patients with no history of borderline PD, the rate of MOH declined from 62.0% to 37.0%.26

We acknowledge that this study is vulnerable to several criticisms. As a single-group retrospective outcome study, with no randomized, controlled comparison treatment, it is not possible to rule out other factors that might have contributed to the observed outcomes. Note, however, that these patients represent treatment failures from a national referral base (43 states and Canada), in some cases from respected experts in headache management, who had failed multiple other treatments including unsuccessful efforts to withdraw from overused medication on an outpatient basis.

One could argue that spending 13 days in an institution where meals are provided, rest is assured, and outside stressors are minimized, would at least partially serve to alleviate any significant biopsychosocial contributions to CDH, or at least alter perceptions of pain and the limitations caused by it. While there may be some truth to this argument, note that many of these patients had received previous, less comprehensive inpatient treatment in other settings – primarily based on rest, withdrawal from daily stressors, and intravenous medication – with minimal improvement. In our program, patients were required to be up, dressed, active, and participate in normal daily activities (eg, make their own beds), and attend groups. In addition, patients confronted relevant stressors and life concerns on an ongoing basis in their psychotherapeutic sessions, conjoint family sessions when appropriate, and daily physician rounds.

The focus on program completers may appear biased, given the current standard in therapeutic drug trials to base statistics on the intent-to-treat population (including dropouts, typically classified as unimproved). However, only 6% of the admitted patients were discharged prematurely in contrast to the relatively high dropout rates in many therapeutic trials. In contrast to a controlled drug trial, an efficacy study of a treatment program on targeted symptoms is more relevant when evaluating those who actually completed the program, although the number of dropouts has relevance.

The evaluation of the primary outcome measures reflected the input and judgment of the primary attending physician (J.R.S.) as well as the clinical team. Ratings of outcome, even by consensus of several treatment providers, may appear too vulnerable to bias. However, we believe that the composite improvement ratings as used here may be the most valid measure for a subjective illness such as headache, taking into account a wide range of data from several different treating sources as well as the impressions of the patient, which were generated through 4-time-a-day pain ratings. The team observed patients' behavior as well as capacity to participate and function in program events throughout a LOS of up to 2 weeks. In contrast, some of the more standardized measures, such as the Migraine Disability Assessment score (MIDAS), are most appropriate for longer periods of observation (eg, 3 months).33 The measures of successful outcome reported here were selected to address the principal goals of referring physicians and patients: rapid reduction in pain, successful drug withdrawal in cases of MOH, determination of effective abortive/rescue medication, identification of behavioral factors that can enhance or complicate treatment responsiveness, and recommendations for additional treatment goals post discharge.

Inpatient headache treatment can vary significantly from one setting to another. In some cases short-term hospitalizations are limited to intravenous infusion protocols only.6 For some of our patients, previous hospital treatment consisted of intravenous opioids. In contrast, the inpatient program described here is comprehensive and multidisciplinary, including a wide range of individually tailored medical and behavioral interventions, as well as a therapeutic group milieu. Any generalizability of the results would appear limited to similar comprehensive, multidisciplinary programs.

This study provides only short-term outcome data at discharge. Maintenance of improvement over time is the most important goal, but long-term sustained improvement requires initial clinical redirection. It also requires identification and ultimately control of a variety of several clinical variables, including compliance with treatment recommendations and restrictions, the quality of follow-up care, and home or work environment factors, among many others.

For a chronic disorder such as primary headaches, an essential first step is the transition from intractability to some level of improvement. This study assessed the effectiveness of the critical first step in that process.