|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
© 1999 American Society for Clinical Oncology Implementing Guidelines for Cancer Pain Management: Results of a Randomized Controlled Clinical TrialFrom the Swedish Medical Center, Seattle, WA. Address reprint requests to Stuart Du Pen, MD, 1221 Madison, No. 410, Seattle, WA 98104; Email stuart.dupen{at}painconsult.com
PURPOSE: Pain and symptom management is an integral part of the clinical practice of oncology. A number of guidelines have been developed to assist the clinician in optimizing comfort care. We implemented clinical guidelines for cancer pain management in the community setting and evaluated whether these guidelines improved care. PATIENTS AND METHODS: Eighty-one cancer patients, aged 37 to 76 years, were enrolled onto a prospective, longitudinal, randomized controlled study from the outpatient clinic settings of 26 western Washington–area medical oncologists. A multilevel treatment algorithm based on the Agency for Health Care Policy and Research Guidelines for Cancer Pain Management was compared with standard-practice (control) pain and symptom management therapies used by community oncologists. The primary outcome of interest was pain (Brief Pain Inventory); secondary outcomes of interest were all other symptoms (Memorial Symptom Assessment Scale) and quality of life (Functional Assessment of Cancer Therapy Scale). RESULTS: Patients randomized to the pain algorithm group achieved a statistically significant reduction in usual pain intensity, measured as slope scores, when compared with standard community practice (P < .02). Concurrent chemotherapy and patient adherence to treatment were significant mediators of worst pain. There were no significant differences in other symptoms or quality of life between the two treatment groups. CONCLUSION: This guideline implementation study supports the use of algorithmic decision making in the management of cancer pain. These findings suggest that comprehensive pain assessment and evidence-based analgesic decision-making processes do enhance usual pain outcomes.
NUMEROUS ORGANIZATIONS, INCLUDING ACADEMIC, private sector, and government agencies, have developed guidelines, standards, and "templates" for pain and symptom management in oncology and end-of-life care. Recent medicolegal controversies surrounding physician-assisted suicide have increased calls for practice guidelines for palliative care. Practical aspects of implementation of guidelines for pain management, and subsequent outcome studies measuring guideline effects on practice, have been essentially nonexistent. Measuring outcomes from any kind of practice guideline is a relatively new field. The earliest controlled trials of guideline implementation were monitored simply by noting whether an order was written, a test result was obtained, or a drug was prescribed.1,2 The most sophisticated systems use specific therapy recommendations built into physicians' computerized order-writing workstations.3-5 In general, guidelines are a reflection of best practice for a population of patients rather than individual members of the population. An algorithm for cancer patient treatment was developed by the Pain Management Service at the Swedish Medical Center in Seattle, WA. The cancer pain algorithm is a decision-tree model based on the Agency for Health Care Policy and Research (AHCPR) Guidelines for Cancer Pain Management.6 The guidelines provide an evidence-based foundation for treatment recommendations, but they do not provide explicit treatment protocols that can be clinically tested. The cancer pain algorithm is a practical interpretation that, as accurately as possible, represents a working model of the AHCPR guidelines. The basic decision-process template of the cancer pain algorithm is pain assessment, analgesic drug choice decisions, and reassessment in a reiterative cycle design that anchors around the balance of analgesic efficacy versus toxicity. Assessment incorporates a differential diagnosis of the pain and reinforces the need to use palliative antitumor therapy, including chemotherapy, hormonal therapy, and radiation therapy as appropriate. The analgesic drug decision-making intervention is a longitudinal process. There is no single intervention, but rather the algorithm process of assessment, treatment decisions, and reassessment is continuous over time. The algorithm is designed to be a process intervention used for the full duration of pain treatment rather than a one-time treatment approach. The process is operationalized with a set of tools, starting from the initial assessment. A clinic flow sheet is used to document the intensity of the pain, note the presence of any neuropathic pain character, and note the presence of any pain- or analgesic-related side effects in much the same way that oncology flow sheets document hematologic trends and chemotherapy dose adjustments. A bulleted set of analgesic guiding principles for opioids, nonsteroidal anti-inflammatory drugs, tricyclic antidepressants, and anticonvulsants is available for the oncology clinic staff for reference. Each drug category guideline contains supplemental information on indications for the drug, specific agents that are recommended, and practical drug administration guidelines that include dosage and titration, route-specific information, and special attention to pain crisis intervention. The algorithm decision tree directs the oncologist/oncology nurse to comprehensive side effect protocols, equianalgesic conversion charts, and a primer for intractable pain. Side effect protocols cover constipation, nausea, dry mouth, sedation, myoclonus, and gastrointestinal distress. Equianalgesic charts are taken from the AHCPR guidelines. A flow sheet for each patient's chart was created to monitor significant pain and symptom indicators against their analgesic therapy. All of these tools were designed to be similar in concept to an antitumor therapy protocol, with the goal of maximum ease of use in the outpatient oncology setting. We designed a two-part study to test the guideline in clinical practice, instruct community oncologists and nurses in the algorithm method, and then retest the community-driven algorithm. The purpose of the first part of the study was to compare the cancer pain algorithm to standard practice, which was defined as "current pain assessment and analgesic drug choice decisions made by community oncologists in their daily practice of oncology." The hypothesis of the study was that cancer patients treated as outlined by the guideline-based cancer pain algorithm would report less pain, fewer side effects, and better quality of life than control group patients treated with standard-practice pain management.
Algorithm Development We used Margolis' model of algorithm development7 and relied on previous pain algorithm development done by Cleeland and colleagues.8 Three authors were involved in review and revision of the AHCPR guidelines (S.L.D., A.R.D., and K.S.). Several editions of the working decision-tree model were reviewed and revised by a panel of expert cancer pain clinicians and researchers. Members of the algorithmic development panel were selected because of their involvement in the development and refinement of the AHCPR cancer pain guidelines. Members of the team included Betty Ferrell, PhD, RN, FAAN (City of Hope Medical Center, Duarte, CA), Charles Cleeland, PhD (Pain Research Group, MD Anderson Cancer Center, Houston, TX), Stuart Grossman, MD, and Vivian Sheidler, RN, MSN (both from the Department of Neuro-Oncology, Johns Hopkins Medical Center, Baltimore, MD), and Michael Levy, MD, and Pam Kedziera, RN, MSN (both from the Department of Supportive Care, Fox Chase Cancer Center, Philadelphia, PA). Two practicing physician/nurse teams were chosen specifically because of their ability to evaluate the feasibility of utilizing the algorithm in a working oncology practice setting. Best consensus was obtained through written review and verbal consultations.
Methods A total of 96 cancer patients were randomized to the 3-month study, with 81 completing at least the 1-month assessment. This cutoff point of 1 month was selected before the start of the study and was believed to meet minimum criteria for an adequate test of the efficacy of the longitudinal intervention. The 81 patients for whom outcome data are reported, and the 15 patients who dropped out of the study before 1 month, are described in Table 1.
Patients were referred to the study by their participating oncologists. A screening telephone call was made to describe the study and to verify that the patient had pain at a level 3 or higher on a 0-to-10 scale. At an initial visit, informed consent and baseline measurements were obtained, and then patients were randomized to either a pain algorithm treatment or standard-practice treatment group. The patients were randomized within referring physicians' practices in permuted blocks such that an approximate balance between treatment arms and the treatment assignment of each patient was not predictable. The investigators reinforced the need for the oncologists to provide primary management of antitumor therapies, including palliative antitumor therapies in patients randomized to either group. Patients were followed longitudinally for 3 months. Measurements were collected at baseline and at 2 weeks, 1 month, 2 months, and 3 months after the start of the intervention. The complete battery of study assessments was carried out on all patients at each time point except the 2-week time point, which was only a partial assessment. The standard-practice group continued to have their pain management implemented by their community oncologists, who used their usual pain management and side effect strategies and documented in their usual fashion. Patients randomized to receive the algorithm intervention had their pain management prescribed by the study physicians and nurse practitioner and implemented by the study nurse for the 3 months of the intervention. The referring oncologists were blinded to specifics of analgesic treatment decision making in those patients randomized to algorithm. This was accomplished by keeping a separate analgesic treatment record that could be unblinded at the discretion of the oncologist on a need-to-know basis. Patients had an initial clinic visit with the pain algorithm physician, at which time the intervention was initiated. The study nurse facilitated the assessment of pain and side effects as outlined by the algorithm and titrated medications under the direction of the algorithm physician, using telephone triage, clinic visits, and home visits. Patients in the algorithm group were instructed regarding their role in the algorithmic process, and the importance of reporting increased and/or unrelieved pain or side effects was stressed. Intervention patients were initially evaluated for etiology of their pain. Pain location and quality were differentially examined against the tumor type and known spread pattern. Any unclear etiology or report of new pain required a diagnostic review by the primary oncologist. Patients were subsequently categorized by pain intensity into mild (levels 1 to 3), moderate (levels 4 to 6), or severe (levels 7 to 10) treatment pathways. Pain intensity directed the algorithm prescriber by instigating speed of response, offering a menu of options, and indicating aggressive dosing and titration of specific agents for patients with severe pain. The pain character represented the second level of algorithmic treatment decision making, where guidelines for treatment of neuropathic pain versus somatic or visceral pain were outlined. Decisions were made in the context of efficacy versus toxicity. Efficacy was defined as a pain score of less than 4 on a 0-to-10 scale. Toxicity was defined by patient report as the presence of bothersome side effects. These definitions were established during algorithm development to provide a standard for when treatment should be initiated or titrated. The algorithm also drove routine reassessment. The most recent pain intensity score determined frequency of contact. The method of contact (ie, telephone call, home visit, or clinic visit) was based on the acuity of the patient's condition. Analgesic dosing and titration, drug changes, and side effect management were implemented by the algorithm physician/nurse as guided by the decision tree. A data collection nurse who recorded outcome data, but was blinded to patient treatment randomization, collected data for both the algorithm and standard-practice groups.
Measures The daily pain diary also included a list of current medications and a daily record of medication taken. Patients completed this measure daily for the 7 days before each assessment time point. Patient satisfaction with pain management was measured on the Pain Treatment Acceptability Scale (PTA),10 which was developed and used by the investigators in a prior clinical trial of cancer pain treatment. The PTA uses mean scores of six items rated from "totally agree" to "totally disagree," such as "I would choose to have this pain treatment again" and "This pain treatment was easy for me to follow." Side effects were measured on the Memorial Symptom Assessment Scale (MSAS),11 and quality of life was measured on the Functional Assessment of Cancer Therapy Scale (FACT).12 Pain treatment adherence was measured on a prescriber and a patient adherence tool developed for the study. The prescriber adherence tool is a chart audit tool that measures how closely the prescribing pattern approximates best practice guidelines in the area of drug choice (ie, strong opioid for severe pain, nonsteroidal anti-inflammatory for bone pain, etc.) and appropriateness of dose frequency and dose adjustment (opioid given around the clock for constant pain, availability of "rescue" opioid in a dose-ratio to the around-the-clock agent, titration of opioid for unrelieved pain, etc.). The patient adherence tool is based on a nurse estimate of "drug ordered" to "drug taken," using the patient diary as a source document along with direct communication with the patient. It provides a measure of how closely the patients adhered to the treatment as prescribed.
Statistical Analysis
Patients in the sample had a mean age of 58 years (range, 37 to 76 years). Almost twice as many women as men participated. This is likely a result of an effort early in the study to recruit breast cancer patients and the relatively low numbers of prostate cancer patients being followed by medical oncologists (v urologists) at our sites. At baseline, patients had an average of five locations in which they identified pain, with a range of two to 15 locations specified. Baseline demographic and descriptive data were similar for patients in the algorithm and standard-practice groups (Table 2). There were no significant differences between the two groups on any of the baseline variables. For usual pain reduction measured by slope scores, the pain algorithm method was statistically superior to pain management in the standard group (t = -2.40, df = 79, P < .02), but it was not superior for worst pain reduction (t = -0.135, P = .2). Patients in the standard group had a decrease initially but finished with slightly higher usual pain scores at the end of the study, whereas algorithm patients experienced a steady downward trend in usual pain scores. The algorithm treatment was the main effect influencing usual pain reduction, even when the two strongest confounders (chemotherapy and patient adherence) were introduced using analysis of covariance techniques. Patients in both groups experienced significant steady downward trends in their worst pain scores. Figure 1 illustrates usual and worst pain by study group over the 3-month study period. There was no statistical difference in percentage of pain relief, pain interference, number of pain locations, pain character, or composite BPI scores when the algorithm intervention was compared with standard pain management. Quality of life as measured on the FACT and side effects as measured on the MSAS were not significant for treatment effect. Patients who received the algorithm pain treatment had significantly better scores on two satisfaction variables, measured on the PTA, when compared with standard-treatment patients at the end of the study. Patients' satisfaction with current pain treatment (t = 2.35, P < .02) and the number of patients who would choose to have similar treatment again (t = 3.22, P < .003) were higher in the intervention group. Improvement in patient satisfaction with pain treatment (slope scores) was correlated with improvement in usual pain (r = -.27, P < .02) more so than improvements in worst pain (r = -.20, P = .07).
Analgesic Use Significantly more adjuvant drugs (nonsteroidal anti-inflammatory drugs, tricyclic antidepressants, and anticonvulsant drugs) were prescribed for algorithm patients (t = 4.281, P < .000). Prescriber guideline adherence to all analgesics (opioids and adjuvants) was 82% for algorithm prescribers and 44% for standard prescribers. Total prescriber adherence to guidelines (opioids and adjuvants) was correlated with both usual pain (r = -.32, P < .005) and worst pain (r = -.24, P < .04) at the end of the study.
Influence of Patient Adherence to Analgesic Therapy
Across treatment groups, both worst pain (measured as slope scores, r = .25, P < .03) and reduction in pain slope scores (as reported in the patients' diaries, r = .31, P < .006) were correlated with patient adherence to analgesic therapy. The slope of usual pain did not reach statistical significance when correlated with patient adherence (r = .16, P = .1). Two-way analysis of variance, using patient adherence as a covariate, indicated a significant confounding effect of nonadherence on worst pain reduction (P < .02), whereas reduction in usual pain was statistically correlated with primary treatment effect (P < .02), despite the introduction of the adherence effect. There was a correlation over time (slope scores) between patient nonadherence and higher levels of distressing symptoms in general on the MSAS (r = -.24, P < .03) and lesser quality of life as measured on the FACT (r = .30, P < .007).
Influence of Chemotherapy There was an equal distribution of patients receiving chemotherapy across the algorithm and standard-treatment groups from baseline through 2 months, with more patients in the standard-treatment group receiving chemotherapy at 3 months (Table 4). Breast cancer patients comprised the largest diagnostic group of patients receiving chemotherapy (54%); lung cancer (13%), prostate cancer (11%), and multiple myeloma (11%) accounted for the majority of other diagnoses. Tests of slope scores for usual and worst pain, covarying for chemotherapy at baseline, indicated no main effect for chemotherapy. As Fig 2 demonstrates, the standard-treatment group seemed to reflect the largest influence of presence or absence of chemotherapy on worst pain. Among the patients in the standard-treatment group, those who received chemotherapy during the early phase of the trial reported significantly lower scores on worst pain than those who did not receive chemotherapy (Table 5). When chemotherapy was factored out, patients in the algorithm group had significantly lower worst pain scores than patients in the standard-treatment group in both early (t = -2.70, P < .008) and late (t = -2.2, P < .04) phases of the study. The severity of neuropathic pain was correlated with type of chemotherapy across all time points (r = -.18, P < .02).
Opioid use was also examined for chemotherapy effect. The 24-hour opioid dose at 1 month was significantly lower in the patients receiving chemotherapy compared with the patients not receiving chemotherapy (mean ± SD, 38 ± 65 mg/d v 96 ± 155 mg/d, respectively; t = 2.1, P < .04). There was no difference between the algorithm and standard-treatment groups in the mean dose of opioid prescribed at 1 month (62 ± 109 mg v 72 ± 132 mg, respectively) (see Table 6). Results of post hoc analysis of the FACT and MSAS data at the 1-month time point are presented in Table 7. When data were analyzed for treatment effect in patients not on chemotherapy, many scores were significantly lower in the algorithm group compared with the standard-treatment group. Notably, nausea was also significantly lower in overall and "no chemo" comparisons by MSAS as well as FACT analysis.
These findings underscore the difficulty of performing pain and symptom management research in the advanced cancer population. The advantage of the algorithmic pain treatment for usual pain reduction was significant in this population of 81 patients, despite important confounding interactions with chemotherapy and patient adherence. Usual pain is arguably the most important of all the pain outcomes because it represents the pain that patients experience "most of the time." We also found that the closer prescribers adhered to guidelines, the more impact that pain treatment had on reduction of usual pain. It is clear that using "best practice" guidelines in this population did have an impact on what patients reported as the pain that they experience most of the time. It is possible that usual pain in the mild- to moderate-intensity range responded well to the more frequently prescribed nonsteroidal anti-inflammatory drugs and/or adjuvant agents used to treat neuropathic pain. Opioid therapy, although not different between the groups in total dose prescribed, was administered more consistently in the algorithm group, which used the preferred around-the-clock dosing schedule. Pain studies have repeatedly shown that preventing pain is more easily accomplished than treating pain once it is present. The use of adjuvant drugs for cancer pain management is clearly recommended by the AHCPR. Levy13 and others have pointed out the importance of the "opioid-sparing effect" of optimizing adjuvant analgesic therapies. Theoretically, the degree of analgesic efficacy afforded by the adjuvant drugs can decrease the amount of opioid required and thus decrease the incidence of opioid-related side effects. Despite more aggressive use of adjuvants in the algorithm group, we were unable to show a reduction in the opioid dosing or "opioid-sparing" effect or a reduction in opioid-related side effect outcomes as measured on the MSAS composite score. These data suggest that either adjuvants do not have an opioid-sparing effect or the total opioid dose in the standard group needs to be higher to match the efficacy in usual pain demonstrated by the opioid/adjuvant combination intervention. If analgesic efficacy had been equal in the two groups, higher opioid dosing in the standard-treatment group would likely have been needed if adjuvants were not used, and this could have resulted in more side effects. Another possible explanation is that the algorithm impact on side effects is lost in the global symptoms and side effects associated with chemotherapy and advancing cancer in general. A larger sample is needed to fully measure these interactions. The relationship between what patients report as the worst pain they experience and their responses to pain treatment versus chemotherapy is particularly unclear. As described on the BPI, worst pain is a reflection of the patient's worst level of pain intensity over the past 24 hours. Worst pain was the variable associated with significant confounding by the pain-reducing effects of chemotherapy. The pain-reducing impact of antitumor therapy on chemotherapy-responsive tumors may be seen primarily in patients' reports of a reduction in worst pain. Some patients clearly reported more neuropathic pain character in the days after chemotherapy, particularly chemotherapy with taxol. Subsequent pain reports may reflect a significant drop in pain coming after a temporary increase in worst pain.
Another issue is the terminology of "worst" pain. Is worst pain as defined on the Brief Pain Inventory capturing "breakthrough" pain? The AHCPR guidelines support treating breakthrough pain with additional doses of prn medications but do not give a recommended method of calculating doses.6 Experts have published recommended ranges of rescue opioid doses that are a ratio of rescue to baseline opioid doses. These recommended rescue doses range from 4% to 30% of the 24-hour opioid dose.14-16 For this study, we chose 10% to 30% of the 24-hour opioid dose as the appropriate range for a rescue dose of opioid. Although we adhered rigorously to this ratio of prn to around-the-clock opioid, there was no advantage for worst pain with the intervention. Interestingly, a recent study examining the rescue-dosing relationship of oral transmucosal fentanyl citrate failed to demonstrate a relationship between the 24-hour dose of baseline opioid and the dose of supplemental opioid required to produce analgesia.17 Some patients had fairly low-level constant pain but severe "spikes" of breakthrough pain, whereas others had relatively high levels of constant pain with relatively small "blips" of breakthrough pain. These findings suggest that, at least for oral transmucosal fentanyl citrate, rescue doses should be titrated to efficacy independently of the 24-hour opioid dose. A clearer terminology for evaluating efficacy of pain treatment might include "baseline" or "persistent" pain rather than usual pain, and breakthrough or "episodic" pain rather than worst pain. Some patients indicated to the investigators that there was a level of pain that was tolerable to them. It is possible that once the baseline or usual level of pain is controlled, episodic or brief increases in pain level are tolerated by some patients in lieu of taking more drug. This relationship between worst pain and nonadherence may also be related to analgesic toxicity. Patients may choose pain over side effects. It is notable, though, that those who had more symptom distress and lower quality-of-life indicators were clearly less adherent to analgesic therapy. These patients may be particularly vulnerable to poor symptom management through lack of understanding of the pain treatment plan or inability, for other reasons, to comply with treatment plans. Further research into reasons these patients do not adhere to treatment is essential if we are to solve the puzzle of why patients receive inadequate pain relief when adequate treatments are available. The phenomenon of patient adherence to analgesic therapy was explored in this study, but the extent of nonadherence was not anticipated. Study nurses, working with patients in their homes, were continuously confronted with patients' refusal to take prescribed treatments even when pain was not well controlled and there were no side effects. Most often, when patients stopped taking prescribed medication, they no longer discussed pain or its treatment with their physicians. Barriers to successful cancer pain management are multifactorial and have been well described in the literature.18-20 Patient factors include fear of addiction, association of pain escalation with disease progression, fear of side effects of medication, fear of disapproval of family and friends, and reluctance to report pain to the doctor. Adherence, or compliance, to treatment represents a broad and complex area explored in many medically treated conditions.21-25 Factors identified that contribute to nonadherence include lack of perceived efficacy, incidence of side effects, polypharmacy, misinformation, and cost. This study found patient adherence to analgesic and side effect regimens to be problematic across the board, with no definitive way to categorize or define the issues. In the next phase of this research, two new components were added to address patient adherence. The Patient Barriers Survey26 was added to the assessment tools, and a post-study interview with subsequent qualitative analysis will be conducted to examine sources of analgesic adherence further. One issue that became apparent during the study is that although oncologists continuously monitor pain and side effects, there is often no documentation or written follow-up that this is occurring. Physicians participating in the study related that time constraints in today's health care system create a barrier to expanded documentation on symptom management. However, a structured assessment template can be designed for the medical record that combines patient self-report (ie, waiting room checklist), nurse triage (ie, clinic flow sheet), and a concise physician treatment plan. In the algorithm group, the nurse performed the majority of routine pain and symptom assessment. The nurse queried the patients on location, intensity, and character of pain and charted their responses in the clinical progress note. A checklist of side effects, including constipation, oversedation, and nausea, was also covered as a part of the clinical assessment. If the patient's pain was well controlled, the nurse reinforced to the patient the need to report future changes in pain and/or the presence of bothersome side effects. If a patient had new pain or increasing pain, that information received a priority delivery to the physician. A patient education booklet covering a variety of side effects, from constipation to dysphoria, was available for patients and their caregivers. These tools were designed to streamline the time required from the physician as much as possible. Valid questions are being raised about how to optimize symptom outcomes in time- and resource-restricted clinical practice. Does adherence to guidelines improve clinical practice? Clinicians agree that blindly following guidelines will obscure critical individual differences and result in poor care. Guidelines are only expected to apply to 60% to 95% of relevant cases.27 In this study, prescriber compliance with optimized analgesic guidelines was highly correlated with outcomes, yet intervention prescribers chose to follow the guidelines only 82% of the time. Perhaps more significantly, even when optimized therapy was available, patients often chose not to comply with recommended therapy. These issues of adherence cannot be separated from evaluations of guideline efficacy. A modest guideline that allows flexibility for the physician and accounts for patient readiness to comply with therapy may prove more efficacious than an elegant practice guideline which no one uses. This study demonstrated that cancer patients treated with a pain algorithm process achieved a statistically significant advantage in usual pain levels over time when compared with a control group representing standard pain management practices in the community. The pain algorithm is an implementation strategy based on the AHCPR Guidelines for Cancer Pain Management. Algorithmic approaches to clinical guideline implementation have been purported to serve the basic goals of quality improvement and cost control.28 Although treatment recommendations for palliative care usually emphasize efficacy, clinicians in practice must consider integration of palliation with other ongoing medical care, such as chemotherapy, with confounding from other symptom treatment needs, with costs, with competing health priorities, and with the perceived benefit of the intervention by individual patients. This study demonstrates that guidelines can improve pain treatment efficacy, but equally important, this study demonstrates the complex issues than can confound both implementation of a guideline and measurement of outcomes in the real world of clinical practice.
Supported by grant no. CA-64877 from the National Cancer Institute and by a grant from the William Gates Foundation. K.S. was supported by grant nos. CA-63030 and CA-68139 from the National Cancer Institute. We acknowledge the expert assistance of Betty Ferrell, PhD, Stuart Grossman, MD, Raymond Houde, PhD, Pam Kedziera, RN, Michael Levy, MD, and Vivian Sheidler, RN, in the development and refinement of the algorithm. We also acknowledge the valuable contributions of Lyn Sullivan Lee, RN, Rhonda Niles, RN, Judy Kornell, RN, Caitie Conley, and Whitney Roose.
The contents of this study are solely the responsibility of the authors and do not necessarily represent the official views of the National Cancer Institute or the William Gates Foundation.
1. McPhee SJ, Bird JA, Fordham D, et al: Promoting cancer prevention activities by primary care physicians: Results of a randomized, controlled trial. JAMA 266:538-544, 1991 2. Therney WW, Hut SL, McDonald CJ: Delayed feedback of physician performance versus immediate reminders to perform preventive care: Effects of physician compliance. Med Care 24:659-665, 1986[Medline] 3. McDonald CJ, Hut SL, Smith DM, et al: Reminders to physicians from an introspective computer medical record: A two-year randomized trial. Ann Intern Med 100:130-138, 1984 4. Barnett GO, Winickoff RN, Morgan MM, et al: A computer-based monitoring system for follow-up of elevated blood pressure. Med Care 21:400-409, 1983[Medline]
5.
Sittig DR, Stead WW: Computer-based physician order entry: The state of the art. J Am Med Inform Assoc 1:108-123, 1994 6. Jacox A, Carr DB, Payne R, et al: Management of Cancer Pain Clinical Practice Guideline No. 9 (AHCPR Publication No. 94-0592). Rockville, MD, Agency for Health Care Policy and Research, U.S. Department of Health and Human Services, Public Health Service, March 1994.
7.
Margolis CZ: Uses of clinical algorithms. JAMA 249:627-632, 1983 8. Cleeland CS, Rotondi A, Brechner T, et al: A model for the treatment of cancer pain. J Pain Symptom Manage 1:209-215, 1986[Medline] 9. Daut RL, Cleeland CS, Flanery RC: Development of the Wisconsin Brief Pain Questionnaire to assess pain in cancer and other diseases. Pain 17:197-210, 1983[Medline] 10. Syrjala KS, Abrams JR, Cowan J, et al: Is educating patients and families the route to relieving cancer pain? Symposium, 8th World Congress on Pain. Seattle, WA, IASP Press, 1996 (abstr 167) 11. Portenoy RK, Thaler HT, Kornblith AB, et al: The Memorial Symptom Assessment Scale: An instrument for the evaluation of symptom prevalence, characteristics and distress. Eur J Cancer 30A:1526-1536, 1994
12.
Cella DF, Tulsky DS, Gray G, et al: The Functional Assessment of Cancer Therapy Scale: Development and validation of the general measure. J Clin Oncol 11:570-579, 1993
13.
Levy MH: Pharmacologic treatment of cancer pain. N Engl J Med 335:1124-1132, 1996 14. Cherny NI, Portenoy RK: Cancer pain management. Cancer 72:3393-3415, 1993[Medline] 15. Levy MH: Pharmacologic treatment of cancer pain. N Engl J Med 335:1124-1132, 1996 16. Cleary JF: Pharmacokinetic and pharmacodynamic issues in the treatment of breakthrough pain. Semin Oncol 24:S16-19, 1993 17. Coluzzi P: A titration study of oral transmucosal fentanyl citrate for breakthrough pain in cancer patients. Proc Am Soc Clin Oncol 16:41A, 1994 (abstr) 18. Ward SE, Goldberg N, Miller-McCauley V, et al: Patient-related barriers to management of cancer pain. Pain 52:319-324, 1993[Medline] 19. Bressler LR, Geraci MC, Schatz BS: Misperceptions and inadequate pain management in cancer patients. Drug Intell Clin Pharm 25:1225-30, 1991 20. Hill CS: The barriers to adequate pain management with opioid analgesics. Semin Oncol 20:1-5, 1993[Medline] 21. Berndt S, Maier C, Schutz H: Polymedication and medication compliance in patients with chronic non-malignant pain. Pain 52:331-339, 1993[Medline] 22. D'Souza W, Burgess C, Ayson M, et al: Trial of a "credit card" asthma self-management plan in a high risk group of patients with asthma. J Allergy Clin Immunol 97:1085-1092, 1996[Medline] 23. Hanchak NA, Patel M, Berlin J, et al: Patient misunderstanding of dosing instructions. J Gen Intern Med 11:325-238, 1996[Medline] 24. Melnikow J, Kiefe C: Patient compliance and medical research: Issues in methodology. J Gen Int Med 9:96-105, 1994[Medline] 25. Sclar DA, Tartaglione TA, Fine MJ: Overview of issues related to medical compliance with implications for the outpatient management of infectious diseases. Infect Agents Dis 3:266-273, 1994[Medline] 26. Ward SE, Hernandez L: Patient-related barriers to management of cancer pain in Puerto Rico. Pain 58:233-238, 1994[Medline]
27.
Eddy DM: Designing a practice policy: Standards, guidelines, and options. JAMA 263:3077-82, 1990 28. Murrey K, Gottlieb L, Schoenbaum SC: Implementing clinical guidelines: A quality management approach to reminder systems. Qual Rev Bull 12:423-433, 1992 Submitted March 9, 1998; accepted September 9, 1998.
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 1999 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
2 tests. Planned analysis, using t tests, included an examination of the difference scores from baseline to early (mean of 2 weeks and 1 month) and baseline to late (mean of 2 months and 3 months). Slope analysis of pain scores over the 3-month time frame were also planned to examine the trend overall and the rate at which intervention-based changes in outcomes developed. Slope scores are particularly helpful when an intervention is a process that occurs over time rather than a single or short-term intervention. Slope analysis easily accommodates the missing values expected in the advancing cancer population. 
