This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via HighWire Citing Articles via Web of Science (6) Citing Articles via Google Scholar Google Scholar Articles by Knab, J. H. Articles by Weinger, M. B. Search for Related Content PubMed PubMed Citation Articles by Knab, J. H. Articles by Weinger, M. B. Related Collections Pain Technology

---

PAIN MEDICINE

The Use of a Computer-Based Decision Support System Facilitates Primary Care Physicians’ Management of Chronic Pain

John H. Knab, MD^*, Mark S. Wallace, MD, R. Lee Wagner, MD^*, John Tsoukatos, PhD^||, and Matthew B. Weinger, MD

*Anesthesia Services Medical Group; Departments of Anesthesiology and Pain Medicine, University of California, San Diego; VA San Diego Healthcare System; and ||NovaTelligence, Inc., San Diego, California

Address correspondence and reprint requests to Matthew B. Weinger, MD, VA Medical Center (125), 3350 La Jolla Village Dr., San Diego, CA 92161–5085. Address e-mail to mweinger{at}ucsd.edu

	Abstract

Top Abstract Introduction Methods Results Discussion References

 
We tested whether computer-based decision support (CBDS) could enhance the ability of primary care physicians (PCPs) to manage chronic pain. Structured summaries were generated for 50 chronic pain patients referred by PCPs to a pain clinic. A pain specialist used a decision support system to determine appropriate pain therapy and sent letters to the referring physicians outlining these recommendations. Separately, five board-certified PCPs used a CBDS system to "treat" the 50 cases. A successful outcome was defined as one in which new or adjusted therapies recommended by the software were acceptable to the PCPs (i.e., they would have prescribed it to the patient in actual practice). Two pain specialists reviewed the PCPs’ outcomes and assigned medical appropriateness scores (0 = totally inappropriate to 10 = totally appropriate). One year later, the hospital database provided information on how the actual patients’ pain was managed and the number of patients re-referred by their PCP to the pain clinic. On the basis of CBDS recommendations, the PCP subjects "prescribed" additional pain therapy in 213 of 250 evaluations (85%), with a medical appropriateness score of 5.5 ± 0.1. Only 25% of these chronic pain patients were subsequently re-referred to the pain clinic within 1 yr. The use of a CBDS system may improve the ability of PCPs to manage chronic pain and may also facilitate screening of consults to optimize specialist utilization.

IMPLICATIONS: The use of a computer-based decision support system may improve the ability of primary care physicians to manage chronic pain and may also facilitate screening of consults to optimize specialist utilization.

	Introduction

Top Abstract Introduction Methods Results Discussion References

 
As many as 80 million Americans experience chronic pain. It is estimated that the annual US expenditures associated with chronic pain amount to $90 billion, and pain accounts for approximately 25% of all sick days taken in the United States (1). In the managed care setting, the chronic pain patient consumes significantly more resources than comparison patients matched for age, sex, and enrollment status (2). The importance of more effective pain management is magnified when one considers the additional costs associated with failed care. These include drug and treatment side effects, wasted expenditure of health care resources, and loss of faith by patients in the health care system. At the time this study was initiated, the VA San Diego Healthcare System (VASDHS) Pain Clinic had a backlog of more than 100 patient referrals, primarily for the medical management of chronic pain. Inadequate resources prevented this backlog from being addressed expeditiously.

Many health care providers do not properly manage pain (3). Pain is often undertreated (4–8), perhaps because of a lack of appreciation for the patient’s degree of suffering or concern about the risks of side effects of potentially effective treatment modalities. Suboptimal pain management may also stem from failure to use appropriate adjuvant drugs, inappropriate dosage regimens, or failure to prevent or treat drug side effects (6).

One of the major barriers to progress in pain management is primary care physicians’ (PCPs’) lack of clinical knowledge about pain diagnosis and therapy (9,10). Even knowledgeable practitioners find chronic pain management challenging. Chronic pain is rarely cured and, instead, both physician and patient must direct their efforts toward enhanced daily function and an optimal quality of life, despite pain. Many pain treatment plans incorporate therapeutic trial and error whereby a series of medications are tried until an effective approach is identified. This necessitates frequent physician visits, reevaluations, and medication adjustments.

This study grew out of the belief that PCPs could more effectively manage patients with chronic pain, thereby reducing specialist referrals, if they were provided with appropriate knowledge and skills. There are a variety of approaches to improving physicians’ knowledge and skills, including the use of educational materials, formal conferences, local seminars, focused educational interventions, and computer-based instruction. Studies have demonstrated that the clinical information needs of the clinic-based practitioner are not met by traditional medical textbooks or journals (11). These written educational materials are of limited value as a dynamic information tool because it is difficult to locate and apply their contents to a specific patient’s current problem. Traditional continuing medical education activities do not readily result in altered physician practice behavior, particularly with regard to chronic and cancer pain, perhaps because of significant fears and attitudes regarding the consequences of treatment (7).

In contrast, computer-based decision support (CBDS) can improve the quality of clinical practice (12,13). Advantages of CBDS include the ability to influence medical decisions at the point of care, improved access to the full range of available patient data, relevant clinical knowledge, the transformation of accepted clinical guidelines into dynamic and usable tools for managing specific patients in real time, and the improved collection of patient care information to track outcomes. Data from several studies suggest that CBDS enhances clinical performance, particularly with respect to the selection of appropriate patient therapy (14–16). This study was designed to test the hypothesis that CBDS could allow PCPs to more effectively manage patients with chronic pain.

	Methods

Top Abstract Introduction Methods Results Discussion References

 
After approval of the UCSD Committee on Human Subjects, 100 consecutive referrals to the VASDHS Pain Clinic during the period between March and October of 1997 were studied. Two complementary experiments were conducted in parallel (Fig. 1). Because the patients used in this study had already been referred to the VASDHS Pain Clinic, they had already "failed" primary care management. Thus, additional appropriate management by a comparable PCP as a consequence of CBDS use would suggest that this approach could potentially reduce the incidence of pain specialist referrals by PCPs in actual clinical practice.

View larger version (63K): [in this window] [in a new window]   Figure 1. Flow chart describing the study including the two parallel experimental limbs and the dependent measures of interest. CBDS = computer-based decision support; UCPCP = University of California Primary Care Physician; VAPCP = VA primary care physician; CPS = community pain specialist.

It was important to first assess the usability of the abstracted summaries for clinical decision-making. Therefore, one of the authors (MSW), a pain specialist who did not participate in the abstraction process, reviewed the case summaries for clinical consistency and completeness. Abstracts with incomplete or confusing data were returned to the abstractors for completion or clarification, where possible (if the medical record data were available). A CBDS system, the Pain Management Advisor^TM (PMA^TM version 1.1; NovaTelligence, Inc., San Diego, CA), was then used to generate specific treatment recommendations for each patient. If the system asked about a clinical sign or symptom that was not noted on the case summary, then the reviewer assumed that this sign or symptom was absent. For each case, the reviewer determined whether an acceptable treatment regimen could be obtained by using PMA.

Because of the significant backlog in the VASDHS Pain Clinic at the time, MSW sent a letter to each patient’s referring VA physician that provided recommendations for managing the patient’s pain. The letter stated that a chart review had been performed on the patient and that CBDS software had been used to generate specific pain management recommendations, and then the letter outlined the actual recommendations produced by the PMA. The letter asked the referring physicians to follow the recommended treatment plan before re-referring the patient to the Pain Clinic. The letter also stated that a pain specialist was available to answer questions or provide clinical advice.

Medical CBDS systems use artificial intelligence techniques to allow computers to generate, process, or interpret clinical knowledge (17–19). The PMA used a series of rule-based (i.e., IF-THEN) decision algorithms derived from the expert knowledge of pain specialists. The PMA design goal was to enhance the ability of PCPs to manage patients with complex chronic pain problems. Although the PMA emphasizes medical management (including pharmacologic, nonpharmacologic, physical, and psychosocial modalities), invasive procedures or specialist referrals are recommended when they are clearly indicated or when more conservative therapies have failed. The clinical content of the PMA was based on modifications to published pain algorithms (20) and is consonant with the clinical practice of the Director of the VASDHS (MSW) to which the PCPs referred their patients.

This study was conducted by using PMA version 1.1 running on Pentium-based personal computers that used Windows 95^TM (Microsoft Corp., Redmond, WA). The PMA, written in Microsoft Visual Basic^TM 5.0, ran as an expert application in XpertRule^TM (Atar Software, Leigh, England). One Microsoft Access^TM database contained the algorithms, and a separate database was used to store a record of the user’s interaction with the system and the final treatment choices. Microsoft Help Utility^TM was used to provide detailed explanations about every medical term, concept, diagnostic query, and advised therapies.

After 1 of 20 specific pain algorithms (Table 1) is selected, the PMA guides the user through a series of questions to refine the diagnosis and determine the most appropriate therapy. PMA uses a conservative stepwise approach to pain management. Recommendations are presented as a prioritized list of therapeutic options. At each step, the physician can ask PMA for explanations, therapeutic rationale, and therapy guidelines (e.g., dosing, side effects, drug interactions, etc.). If the physician is dissatisfied with PMA’s initial advice (perhaps because the patient has already been tried on and failed the therapies recommended), then the next level of therapy can be displayed for consideration. Only when all reasonable therapeutic options have been exhausted does PMA recommend pain specialist referral.

The PCPs were instructed to use the PMA for each case to attempt to arrive at a therapeutic plan that, given the clinical information provided, they felt was "medically acceptable" (i.e., they would have been comfortable prescribing it for that patient if they had seen the patient in their actual clinical practice). The PCPs were permitted to "write in" other therapies that did not appear as options in the PMA. They were instructed to assume that a specific clinical sign or symptom was absent if it was not noted specifically on the case summary.

A "successful outcome" was defined as the selection of a new or adjusted therapeutic regimen generated by the PMA that was "medically acceptable" ("outcome" is used herein specifically to indicate a definite therapeutic decision by the clinical subject about how they would manage a simulated chronic pain patient; a successful outcome here does not indicate decreases in pain, increased function, or decreased health care utilization). The time spent using the PMA on each case was measured in seconds. Each PCP also ranked the perceived ease of use of the PMA for each case, using a 10-cm visual analog scale, with 1 being "extremely easy" and 10 "very difficult."

Two independent community-based pain specialists (JHK and RLW) reviewed each case summary and generated their own treatment plans. The pain reviewers were instructed to treat the patients as they would in their actual referral practice. The specialists also independently assessed the medical appropriateness of each of the PCPs’ treatment decisions on a 10-cm visual analog scale (0 = totally inappropriate to 10 = totally appropriate). A medically appropriate therapy was defined as one that the specialist reviewer considered to be safe and had some reasonable chance of improving the patient’s pain condition without unduly delaying clearly superior or more definitive therapy.

One year after the letters containing the PMA’s pain management recommendations were sent to the original referring physicians, each patient’s clinic visit and pharmacy profile were searched on the San Diego VA’s computerized patient record system. The end points of interest were whether the patient had been referred subsequently to the VASDHS Pain Clinic for pain management and whether the patient was currently receiving any pain therapies, particularly those recommended by the PMA. The system does not contain information about medical therapy the patient receives outside the VA.

The dependent variables of interest in the PCP’s PMA use experiment were 1) the therapy reached by the PCP while using the PMA on each case, 2) the percentage of cases reviewed in which no therapeutic regimen could be obtained (e.g., referral to a pain specialist), 3) the average time required to use CBDS to arrive at a therapeutic regimen, 4) the ease-of-use rating for each case, and 5) the appropriateness of the PCPs’ therapies as judged by the two independent pain specialists. To ascertain the effect of the PMA-based recommendation letter to the VA PCPs, the dependent variables were the percentage of the original 50 patients re-referred to the VA Pain Clinic by the 1-yr follow-up and the percentage who were currently receiving recommended or nonrecommended pain therapies. As described in Results, data were analyzed by using analysis of variance (ANOVA) followed by Newman-Keuls tests (21), ² tests, or linear regression analysis. Data are presented as mean ± SEM with a P value of 0.05 considered statistically significant.

	Results

Top Abstract Introduction Methods Results Discussion References

 
Of the original 100 consecutive patients referred to the VASDHS Pain Clinic, 20 were specifically referred for a procedural intervention and were excluded. Of the remaining 80 cases for medical management of pain, 52 charts were available for review and abstraction. Two charts provided insufficient data about the patient’s pain condition and were unusable. The academic pain specialist (APS) was able to formulate an acceptable treatment plan for every one of the remaining 50 cases by using only the PMA software recommendations.

The five PCPs used the PMA on all 50 cases. In 6 of 250 cases, the PCPs inadvertently failed to save their resultant decisions in the program, and as a consequence, the algorithms used were also not recorded (coded as "No algorithm identified" in Table 1). However, in four of these cases, the PCPs wrote down their therapeutic choices (three pharmacologic treatments and one pain specialist referral) on a separate data collection form.

In more than half of the cases, all five PCPs used the same algorithm as did the APS, and in 72% of the cases, at least four of the five PCPs used the same algorithm (Table 1). However, the PCPs used a different algorithm than the specialist in 20% of cases. Most often, differences in algorithm chosen were caused by the presence of diffuse pain (e.g., making it more difficult to choose between low back versus neck versus headache), multiple pain sites (e.g., use of the other chronic pain algorithm versus a site-specific algorithm), or the presence of a contributing disease (e.g., herpes zoster, myofascial pain, arthritis, or diabetic neuropathy). However, ² analysis revealed that the use of a different algorithm compared with that chosen by the APS did not affect either success rate (P > 0.05) or appropriateness of therapy (P > 0.05) significantly.

The PCPs prescribed 793 new or modified therapies (3.2 ± 0.1 therapies per patient per PCP). The number of new prescribed therapies ranged from 0 (immediate pain specialist referral) to 10 per PCP per case. Nonpharmacologic therapies were much more often prescribed than pharmacologic therapies (Table 2). Thirty-nine of the prescribed therapies (4.9%) were not directly recommended by the PMA. However, every therapy so prescribed could be found somewhere in the software’s algorithms.

Overall, the two community pain specialists deemed 77.6% and 57.1%, respectively, of the PCPs’ outcomes as more appropriate than inappropriate (i.e., >5 on a 10-point appropriateness scale), with a mean combined appropriateness score of 5.5 ± 0.1 (Table 3). Concurrence between specialists was calculated by dividing the number of medically similar therapies prescribed by both specialists by the total number of therapies prescribed divided by two. The two pain specialists concurred in only 28 cases (11.7%) in their assessment that the prescribed treatment was more inappropriate than appropriate (i.e., <5). There were five cases in which both reviewers believed that the PCP’s therapy was totally inappropriate. In two of these cases, which were both neck pain cases, both reviewers believed that the patient should have been immediately referred for interventional procedures. Both reviewers were also highly critical of one PCP’s therapy of two back pain cases, although the specialists differed in how they would have treated those patients. Reviewer 1 opted to treat for neuropathic pain in both cases, whereas Reviewer 2 chose surgical evaluation in one case and conservative nociceptive pain-oriented therapy in the other. In a diabetic neuropathy case, the choice by one PCP of amitriptyline as the sole new therapy was deemed grossly insufficient by both specialists, who agreed on the use of gabapentin and opioids. In fact, the two specialists concurred in only 37% ± 3% of their proposed treatments of all 50 cases. Reviewer 1 was at least 50% more likely than Reviewer 2 to concur with the therapeutic plan generated by the APS in his postreferral letter (33% ± 2% vs 20% ± 2%; ², P < 0.05).

There were no significant differences in success rate (², P > 0.05), average time spent, or ease of use (last two comparisons with one-way ANOVA, P > 0.05) between the different pain algorithms used (Fig. 2). Not surprisingly, the use of multiple algorithms was associated with longer elapsed times—one-way ANOVA showed a significant main effect (F = 2.8), but only the comparison with herpes zoster attained significance with the Newman-Keuls test.

View larger version (29K): [in this window] [in a new window]   Figure 2. The average (+ SEM) elapsed time per case in minutes () and the perceived ease of use on a 10-cm visual analog scale score (1 = very easy to 10 = very difficult to use; ) for each computer algorithm are both depicted in this histogram. Note that, in 6 cases, no algorithm ("None") was recorded and, in 20 cases, two algorithms ("Multiple") were used for a single case. The use of more than one algorithm was associated with a significantly longer elapsed time. PCP = primary care physician.

Of the 30 patients who had not been re-referred to the pain clinic and for whom some therapy data were available, 21 (70%) were receiving at least one treatment that had been recommended in the pain specialist’s original letter. Ten other patients were receiving few or no pain medications. The PCPs seemed more predisposed to rely on nonsteroidal antiinflammatory drugs and opioid therapy and seemed reluctant to manage their patients with membrane stabilizers despite the APS’s recommendation to do so. Whereas 14 patients were being treated with some kind of nonsteroidal antiinflammatory drug and 14 were receiving a chronic opioid (with 6 receiving both types of pain therapy), only 4 patients were receiving gabapentin or carbamazepine. In contrast, 7 of the 12 patients re-referred to the pain clinic were taking a membrane stabilizer.

	Discussion

Top Abstract Introduction Methods Results Discussion References

 
The goal of this study was to ascertain whether a CBDS might facilitate more cost-effective chronic pain management. The results suggest that the use of CBDS may improve PCPs’ ability to continue to successfully treat patients who otherwise would be referred to a pain specialist for the medical management of chronic pain. The five PCPs used CBDS to generate treatment recommendations that were medically acceptable to them in 85% of the patients studied. In fact, PCPs chose medically appropriate new therapies (on the basis of pain specialist review) in 81% of back and neck pain patients, and these encompassed one half of the cases studied. CBDS may be more effective in actual clinical practice because critical information (e.g., the character of pain) was missing from many of the abstracted charts. The two board-certified community pain specialist reviewers believed that <10% of the cases should be immediately referred to a pain specialist. Considering that all of the study cases had already been referred to the VASDHS Pain Clinic by a PCP, these results could have substantial implications for reducing the cost of caring for pain patients.

There are several explanations for why the PCP’s use of a different algorithm than that chosen by the APS did not affect either success rate or judged appropriateness of therapy. Redundancy was intentionally built into the algorithms so that the choice of the algorithm (e.g., headache instead of neck pain or vice versa) would not unduly or adversely affect the therapy recommended. In addition, the management of neuropathic pain was fairly uniform across all algorithms. Because a significant percentage of the cases had serious chronic pain with a neuropathic component, regardless of the algorithm choice, similar treatment was recommended.

Because pain perception is integrally tied to each individual’s physical and psychological makeup, it is impossible to provide a unitary "cookbook" approach to a particular pain syndrome. Decision trees or algorithms, however, can provide a practical and efficient way of ensuring that a logical and consistent treatment plan is followed. Toward this end, consensus panel-based practice guidelines for chronic and cancer pain have been promulgated by professional societies (22) and government bodies (23,24). Thus, by providing a list of suggested therapies from which the clinician can voluntarily choose, the CBDS broadens PCPs’ management options. We observed frequent incidences of the PCPs "shopping" through the CBDS to ascertain the range of possible diagnoses, clinical issues, and therapies. In this way, this type of CBDS could become a valuable bedside educational tool.

Although average treatment appropriateness scores were similar between the two pain specialist reviewers, there were substantial variability and differences in opinion. For example, one reviewer scored almost twice as many PCP therapies as the other reviewer as being more inappropriate than appropriate. Some differences in appropriateness ratings may be caused by the inherently subjective nature of scoring other peoples’ decisions, as well as by clinical uncertainty and interpretation related to the (occasionally incomplete) abstracted patient descriptions. In addition, the differences may be partially explained by differences in the two pain specialists’ experience, training, and practice styles. Reviewer 1, who had the lower incidence of inappropriate scores, was an anesthesiologist who had performed his pain fellowship under the pain specialist who developed PMA’s algorithms. In contrast, Reviewer 2 had a background in both anesthesiology and internal medicine and was self-trained in pain management. Reviewer 1 also had more experience managing pain patients in a VA setting. Much of the discordance between the two reviewers could be attributed to differences in treatment philosophy, particularly in chronic back and neck pain in elderly patients and in the early indications for sympathetic blocks in reflex sympathetic dystrophy. Thus, these differences may reflect the evolving and sometimes controversial nature of chronic pain management. In the absence of widely accepted detailed consensus standards, the development of CBDS algorithms for chronic pain management may require substantial customization to match the practice style of local pain specialists.

Both in the computer-based simulated cases and in actual practice, the PCPs seemed reluctant to use contemporary coanalgesics, particularly membrane stabilizers, to treat the neuropathic component of chronic pain, despite strong recommendations to do so. However, recent high-profile controlled studies supporting the use of gabapentin in similar settings (25) were published after this study was completed.

There are a number of limitations to the use of clinical CBDS. The knowledge in a rule-based expert system such as the PMA is static; it must be continuously updated to reflect current medical practice (26). Knowledge acquisition from clinical experts is difficult and time consuming. A CBDS system’s knowledge may not account for regional differences in patient populations or clinical practice styles, particularly in a field such as chronic pain management, in which consensus and evidence-based standards are still evolving. However, technological advances may permit CBDS systems to learn and apply new knowledge (18,19,26).

Another impediment is the reluctance of many physicians to use computers during actual patient care. Yet physicians’ concern that their computer use in the clinic room will impair the physician-patient relationship and reduce patient satisfaction is not supported by controlled studies (27). In addition, many physicians may believe that CBDS will be too time consuming. In this study, PMA required approximately five minutes to use. With experienced users in the clinical setting, the time required to obtain a treatment plan could be reduced to only a few minutes.

Medical CBDS has other advantages. Extensive context-sensitive medical help (e.g., full-text explanations, rationales, and references) are available in real time at the point of care. CBDS systems not only generate an electronic record of the patient encounter, but also provide documentation of the physician’s clinical thought processes. Additionally, on future patient visits, the same (or a different) physician can immediately reenter the treatment plan at precisely the place where he or she left off at the last visit. Finally, these systems can automatically generate prescription forms (with the future potential of electronic transmission to the pharmacy of the patient’s choice), as well as written patient instructions and other educational materials.

In a survey (28), family physicians requested the following features of any CBDS system: 1) have a uniform user interface; 2) be available on a hand-held computer; 3) include drug information (e.g., side effects, interactions, contraindications); 4) contain treatment recommendation overviews; 5) provide patient information materials; and 6) have its clinical information updated regularly. Although the mean ease of use of the system used in this study (4.2 of 10) was only modest, an enhanced version of the software has an improved user interface as well as all of these other capabilities.

The PMA was designed for real-time point-of-care use, and it was assumed that the patient would be available to answer relevant clinical questions (e.g., "Does the patient have lacinating pain?" can be answered either "yes" or "no" during CBDS use). However, in this study, both the pain specialists and the PCPs based their therapeutic decisions on the clinical information provided on a single-page case summary sheet. Although, to the extent possible, the summaries accurately reflected the information available in the medical records, both the PCPs and pain specialists complained of missing clinical information that might have affected their therapeutic decisions. In particular, there was poor documentation of the specific character of the patient’s pain (e.g., burning, lancinating, etc.), the nature and severity of associated symptoms (e.g., allodynia, dysesthesia, etc.), and exacerbating factors. To ensure that all of the subjects were analyzing the same case, we had to institute the rule that if information was missing, it was assumed to be absent. If the alternative approach had been taken, when an important clinical sign was missing, the subjects would have had undue discretion over whether they assumed that patient did or didn’t have that sign. This would have led to each case being viewed differently by each subject and thus precluding any comparisons. Because the CBDS explicitly uses specific signs and symptoms to guide pain management, this study design influenced how the system was used. We believe that clinical outcomes when PCPs use a similar CBDS in actual practice would likely be better because the CBDS would prompt the PCP to obtain and consider important additional clinical symptoms, signs, and diagnostic test results.

The number of subjects in this study was small and may not generalize to a larger PCP population. Unfortunately, logistic and economic constraints precluded a larger subject population or the use of community PCPs. This study did not measure true clinical outcomes (e.g., reduction in patients’ pain scores before and after treatment).

Although a net 63% reduction in pain clinic referrals represents appreciable cost savings for a "capitated" health care system such as the VA, this result must be interpreted with caution. For example, it is not known how much the letter to the referring PCPs, with its admission of an appreciable backlog in the Pain Clinic, dissuaded re-referral independently of whether the PCP chose to follow the management recommendations. However, more than half of the nonreferred patients were managed by their PCPs with medications recommended in the letter, and re-referred patients were seen in the Pain Clinic relatively promptly. Although it is possible that some patients failed to be re-referred to the VA Pain Clinic because they sought care outside the VA, all of the patients not lost to follow-up had evidence in their medical records of nonpain-related care in other VA clinics.

The design of this study may have introduced a bias against specialist referral. Although it was not explicitly stated, the PCPs could have surmised that the incidence of pain specialist referral was one of the dependent measures of interest. Because one of the design goals of the PMA was to reduce inappropriate or unnecessary pain specialist referrals, the management algorithms, which are consistent with a multidisciplinary approach to pain management, are medically conservative. However, the incidence of specialist referral by the PCPs was only slightly less than that of the community pain specialists, who have an inherent incentive to advocate referral. Additionally, only 25% of the patients studied were subsequently re-referred to the VASDHS Pain Clinic. This suggests that when provided with appropriate pain therapy knowledge and experience, PCPs were able to manage this patient population with fewer specialist referrals, both under our limited study conditions and in actual clinical practice.

All of the patients in this study had already been managed in "real life" by another PCP and had been referred to the Pain Clinic for medical management of their chronic pain. The results suggest that the use of CBDS could significantly improve the ability of PCPs to manage chronic pain. Additionally, such a decision aid could possibly permit Pain Clinic personnel to prescreen pain consults and provide feedback of standard management algorithms to PCPs before specialist referral. These novel approaches may lead to improved patient and physician satisfaction and lower costs. A much larger prospective controlled clinical trial with a parallel placebo control group will be necessary to validate whether the use of CBDS in chronic pain management will improve quality of care and reduce specialist referrals.

	Acknowledgments

 
Supported, in part, by NovaTelligence, Inc., San Diego, CA.

The technical assistance of Justin Jimenez, Patrick Kennedy, and Chris Chopper was greatly appreciated. Dr. Gery Schulteis assisted in the statistical analysis.

	Footnotes

 
Presented in part at the Society for Technology in Anesthesia Annual Meeting, Tucson, Arizona, January, 1998; at the annual meeting of the American Pain Society, San Diego, California, October, 1998; and at the International Anesthesia Research Society Annual Meeting, Los Angeles, California, March, 1999.

This study grew out of a University industry collaboration between Drs. Weinger and Wallace at the University of California–San Diego and a small San Diego start-up company, NovaTelligence, Inc. Drs. Weinger and Wallace conceived the original idea for the Pain Management Advisor (PMA) and developed its medical content. Besides the authors, no other NovaTelligence, Inc., employee or representatives had an influence on the conduct of the study or the content of the manuscript. Before inception of this study, Drs. Weinger and Wallace were both awarded stock options for 1% equity in NovaTelligence, Inc. Dr. Wallace was to receive a 1.5% royalty on sales of the PMA software but never received any royalty payments. The University of California–San Diego was compensated directly by NovaTelligence, Inc., for some of the time Drs. Weinger and Wallace spent on this project, through a formal, nonclinical consulting agreement that included University of California lawyer-approved clauses for intellectual property, publication rights, and indemnification. Dr. Tsoukatos was a paid employee of NovaTelligence at the time of this study. Drs. Wagner and Knab have no financial relationship with either NovaTelligence or the University of California–San Diego. The five primary care physician subjects, who were University of California–San Diego employees and had no affiliation with NovaTelligence, received modest compensation for their time in accordance with University of California–San Diego IRB regulations and approval.

	References

Top Abstract Introduction Methods Results Discussion References

 

1. Pain management: opportunities in an underserved growth market. Med Pro Month 1997; 7: 305–12.
2. Clause JC, Osterhaus JT. Healthcare resource use and costs associated with migraine in a managed healthcare setting. Ann Pharmacother 1994; 28: 659–64.[Abstract]
3. Mortimer JE, Barlett NL. Assessment of knowledge about cancer pain management by physicians in training. J Pain Symptom Manage 1997; 14: 21–8.[Web of Science][Medline]
4. Breitbart W, Rosenfeld BD, Passik SD, et al. The undertreatment of pain in ambulatory AIDS patients. Pain 1996; 65: 243–9.[Web of Science][Medline]
5. Cleeland CS, Gonin R, Hatfield AK, et al. Pain and its treatment in outpatients with metastatic cancer. N Engl J Med 1994; 330: 592–6.[Abstract/Free Full Text]
6. Jasinski DR. Pharmacologic misinformation that prevents optimum pain control with narcotic analgesics. In: Hill CS Jr, Fields WS, eds. Advances in pain research and therapy. Vol 11. New York: Raven Press, 1989: 139–43.
7. Levin ML, Berry JI, Leiter J. Management of pain in terminally ill patients: physician reports of knowledge, attitudes, and behavior. J Pain Symptom Manage 1998; 15: 27–40.[Web of Science][Medline]
8. Morgan JP. American opiophobia: customary underutilization of opioid analgesics. Adv Alcohol Subst Abuse 1986; 5 (1–2):163–73.
9. Pargeon KL, Hailey BJ. Barriers to effective cancer pain management: a review of the literature. J Pain Symptom Manage 1999; 18: 358–68.[Web of Science][Medline]
10. Weinstein SM, Laux LF, Thornby JI, et al. Physicians’ attitudes toward pain and the use of opioid analgesics: results of a survey from the Texas Cancer Pain Initiative. South Med J 2000; 93: 479–87.[Web of Science][Medline]
11. Covell DG, Uman GC, Manning PR. Information needs in office practice: are they being met? Ann Intern Med 1985; 103: 596–9.
12. Classen DC. Clinical decision support systems to improve clinical practice and quality of care. J Am Med Assoc 1998; 280: 1360–1.[Free Full Text]
13. Garibaldi RA. Computers and the quality of care: a physician’s perspective. N Engl J Med 1998; 338: 259–60.[Free Full Text]
14. Evans RS, Pestotnik SL, Classen DC, et al. A computer-assisted management program for antibiotics and other antiinfective agents. New Engl J Med 1998; 338: 232–8.[Abstract/Free Full Text]
15. Hunt DL, Haynes RB, Hanna SE, Smith K. Effects of computer-based clinical decision support systems on physician performance and patient outcomes: a systematic review. J Am Med Assoc 1998; 280: 1339–46.[Abstract/Free Full Text]
16. Pestotnik SL, Classen DC, Evans RS, Burke JP. Implementing antibiotic practice guidelines through computer-assisted decision support: clinical and financial outcomes. Ann Intern Med 1996; 124: 884–90.[Abstract/Free Full Text]
17. Shortliffe EH. Computer programs to support clinical decision making. JAMA 1987; 258: 61–6.[Abstract/Free Full Text]
18. Deutsch T, Carson E, Ludwig E. Dealing with medical knowledge: computers in clinical decision making. New York: Plenum Press, 1994.
19. Berner ES, Ball MJ. Clinical decision support systems: theory and practice. New York: Springer, 1999.
20. Irving GA, Wallace MS. Pain management for practicing physicians. New York: Churchill Livingstone, 1997.
21. Winer B. Statistical principles in experimental design. New York: McGraw-Hill, 1971.
22. Wilson PR, Caplan RA, Connis RT. Practice guidelines for chronic pain management: a report by the American Society of Anesthesiologists Task Force on Pain Management, Chronic Pain Section. Anesthesiology 1997; 86: 995–1004.[Web of Science][Medline]
23. Bigos SJ, Bowyer O, Braen G, et al. Acute low back problems in adults: assessment and treatment—clinical practice guideline. Quick reference guide No. 14. Rockville, MD: US Department of Health and Human Services, Public Health Service, Agency for Health Care Policy and Research, 1994.
24. Jacox A, Carr DB, Payne R, et al. Management of cancer pain: adults—clinical practice guideline. Quick reference guide No. 9. Rockville, MD: US Department of Health and Human Services, Public Health Service, Agency for Health Care Policy and Research, 1994.
25. Rowbotham M, Harden N, Stacey BPB, Manus-Miller L. Gabapentin for the treatment of postherpetic neuralgia. JAMA 1998; 280: 1837–42.[Abstract/Free Full Text]
26. Hughes CA, Gose E, Roseman D. Overcoming deficiencies of the rule-based medical expert system. Comput Methods Programs Biomed 1990; 32: 63–71.[Web of Science][Medline]
27. Solomon GL, Dechter M. Are patients pleased with computer use in the examination room? J Fam Pract 1995; 41: 241–5.[Web of Science][Medline]
28. Ebell MH, Gaspar DL, Khurana S. Family physicians’ preferences for computerized decision-support hardware and software. J Fam Pract 1997; 45: 137–41.[Web of Science][Medline]

This article has been cited by other articles:

				K. Beissner, C. R Henderson Jr, M. Papaleontiou, Y. Olkhovskaya, J. Wigglesworth, and M. Reid Physical Therapists' Use of Cognitive-Behavioral Therapy for Older Adults With Chronic Pain: A Nationwide Survey Physical Therapy, May 1, 2009; 89(5): 456 - 469. [Abstract] [Full Text] [PDF]

				D. B. Gordon, J. L. Dahl, C. Miaskowski, B. McCarberg, K. H. Todd, J. A. Paice, A. G. Lipman, M. Bookbinder, S. H. Sanders, D. C. Turk, et al. American Pain Society Recommendations for Improving the Quality of Acute and Cancer Pain Management: American Pain Society Quality of Care Task Force Arch Intern Med, July 25, 2005; 165(14): 1574 - 1580. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via HighWire Citing Articles via Web of Science (6) Citing Articles via Google Scholar Google Scholar Articles by Knab, J. H. Articles by Weinger, M. B. Search for Related Content PubMed PubMed Citation Articles by Knab, J. H. Articles by Weinger, M. B. Related Collections Pain Technology