Arthritis & Rheumatism Official Journal of the American College of Rheumatology
INFORMED CONSENT AND THE PRESCRIPTION OF NONSTEROIDAL ANTIINFLAMMATORY DRUGS JEFFREY N. KATZ, LAWREN H. DALTROY, TROYEN A. BRENNAN, and MATTHEW H. LIANG Objective. To examine disclosure of side effects of nonsteroidal antiinflammatory drugs (NSAIDs) and to identify patient- and physician-specific factors associated with greater disclosure. Methods. Forty-six encounters between rheumatologists and new adult outpatients who were prescribed an NSAID they had not been taking prior to the visit were audiotaped. Reviewers coded the NSAID prescribed, specific side effects mentioned, demographic features of patients and physicians, and patient clinical characteristics. Neither patients nor physicians were aware that side effect disclosure was being studied. Results. A mean of 1.7 side effects was mentioned From the Departments of Medicine and Rheumatology! Immunology, Brigham and Women's Hospital, the Robert B. Brigham Multipurpose Arthritis Center, and the Department of Health and Social Behavior, Harvard School of Public Health, Boston, Massachusetts. Supported by NIH grant AR-36308 to Drs. Katz, Daltroy, and Liang. Dr. Katz is recipient of a Postdoctoral Research Fellowship from the Arthritis Foundation. Jeffrey N. Katz, MD, MS: Department of Rheumatology! Immunology, Brigham and Women's Hospital, and the Robert B. Brigham Multipurpose Arthritis Center; Lawren H. Daltroy, DrPH, MPH: Department of Rheumatology!Immunology, Brigham and Women's Hospital, the Robert B. Brigham Multipurpose Arthritis Center, and Department of Health and Social Behavior, Harvard School of Public Health; Troyen A. Brennan, MD, JD, MPH: Department of Medicine, Brigham and Women's Hospital; Matthew H. Liang, MD, MPH: Departments of Medicine and Rheumatology! Immunology, Brigham and Women's Hospital, and the Robert B. Brigham Multipurpose Arthritis Center. Address reprint requests to Jeffrey N. Katz, MD, MS, Department of Rheumatology!Immunology, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115. Submitted for publication December 6, 1991; accepted in revised form April 30, 1992.
per encounter. Epigastric discomfort was mentioned in 72 % of encounters, while other side effects, including hepatic, renal, hematologic, or central nervous system effects, were mentioned in ::515% of encounters. Three factors were identified as independent predictors of less disclosure of side effects: senior clinician (versus less experienced), patient not taking another NSAID immediately prior to the visit, and patient age 7.0) for this analysis. Bleeding tendency was not rated by Fries et al and was designated in this study as a mild effect. Sixteen tapes were reanalyzed to explore the relationship between prior NSAID use and physician experi-
INFORMED CONSENT AND NSAIDs
1259
ence. Half the tapes were of encounters with senior clinicians half with less experienced physicians. For each encounter, reviewers coded whether side effects were first mentioned by the patient or by the physician. Statistical methods. Bivariate relationships between baseline features and the number of side effects disclosed during the encounter were evaluated using nonparametric (Wilcoxon rank sum) tests. Mult!pl~ linear r~gression (15), log-linear analysis (16), and logistic regressIo~ (17) were used to analyze the relationships between baseline featu~es and side effects in multivariate models. The number of SIde effects disclosed served as the continuous dependent variable in the linear regression. In logistic regression and log-linear analyses, the dichotomous outcome variable was whether 2 or more side effects were disclosed, versus 0 or 1. Categorical indicators for patient education level, sex, race, age, and diagnosis, and physician ca~eg?ry (see above) served as independent covariates. A significance level of 0.007 (0.05 divided by 7) was established to minimize the possibility of Type I error due to testing of 7 predictors.
Table 2. Disclosure of possible side effects of specific nonsteroidal antiinflammatory drugs
RESULTS
present analysis. Audiotapes were incomplete on 3 patients, leaving a study sample of 46 patients. Of 28 arthritis center physicians (79%) who were asked to participate in the study, 22 consented to do so. . Subject characteristics. The group of 46 patients was predominantly white and female (Table 1). The patients were divided evenly with respect to several study variables, including age category «40, 40-59, ~60), education level (some college versus none), experience of the physician they saw, disease category, and whether the patient was receiving a different NSAID or no NSAID treatment at the time of the visit. Ten different NSAIDs were prescribed in the 46 encounters, including naproxen (12 patients), piroxicam (8 patients), ibuprofen (7 patients), and diclofenac, tolmetin, meclofenamate, sulindac, indomethacin, aspirin, and salsalate (:54 patients each). The median number of encounters per physician was 3, with a range of 1-5. Disclosure. Disclosure about each of several specific adverse effects is presented in Table 2. While the possibility of epigastric discomfort was disclosed in 72% of encounters, none of the remaining side effects was mentioned in more than 15% of encounters. The total number of side effects mentioned in each encounter ranged from 0 to 5, with a median of 1.5 and a mean of 1.7. In 11 encounters (24%), no side effects were mentioned. The total number of side effects mentioned did not vary significantly across the different NSAIDs (Kruskal-Wallis ¥ = 15.0,9 degrees offreedom [dt], P = 0.09). There was little correlation (r = 0.10) between the number of side effects disclosed and patient satisfaction with the discussion of the therapy prescribed. The relationships between disclosure of specific side effects and their reported frequency and severity
Subject recruitment. In the parent study, we contacted 604 consecutive eligible patients who were planning a first visit to the clinic. Three hundred forty-five (57%) expressed interest in participating in a prospective study that involved audiotaping of2 visits, 2 telephone interviews, completing questionnaires on 3 instances, and random assignment to a counseling intervention or a control group. After exclusions for logistical reasons (scheduling conflicts and missed appointments), 178 patients were enrolled in the study. Of these, 134 were expected to continue their arthritis care at the clinic with the study physician, and data from their initial visits constituted the parent study sample. Forty-nine of these patients were prescribed a new NSAID at the initial office visit and therefore were eligible for the Table 1.
Baseline characteristics of the 46 patients studied Characteristic
Age 7.0) and 8% of possible less severe side effects, suggesting that disclosure of side effects was not influenced by their severity (I = 0.28, 1 df, P > 0.60). Bivariate associations between baseline variables and the number of side effects disclosed were analyzed with the Wilcoxon rank sum test and are presented in Table 3. More side effects were disclosed to patients >40 years old than to younger patients (mean 2.0 versus 1.1). Patient sex and education level were not significantly related to the number of side effects disclosed. More side effects were disclosed to white than to nonwhite patients (mean 1.8 versus 0.8) although, with only 5 nonwhite patients in the sample, the difference did not reach statistical significance (P = 0.11). Patients with either systemic inflammatory or chronic noninflammatory disorders had significantly more side effects disclosed to them than did patients with arthralgias or self-limited disorders (mean 1.8, 2.2, and 1.1 side effects, respectively). The strongest associations were observed for physician experience and whether NSAIDs had been used previously. Less experienced physicians disclosed more side effects than did senior clinicians (mean 2.1 versus 1.2) (although the difference [P = 0.02] did not reach statistical significance based on the stringent standard of P < 0.007 that had been established), and patients taking NSAIDs prior to the visit were told about significantly more side effects than were patients who were not taking NSAIDs prior to the visit (2.3 versus 1.1; P = 0.005). Several multivariate models were used to identify factors independently associated with the total number of side effects disclosed, controlling for all other variables. Stepwise linear regression confirmed that physician experience, patient prior use of NSAIDs, and patient age were associated independently with the extent of disclosure. Log-linear modeling with a bivariate outcome (0--1 side effects versus 2:2) also confirmed the independent predictive value of the same 3 variables. There were cells with 0 patients in the logistic regression and log-linear models, which produced unstable estimates of the model coefficients;
Table 3. Number of side effects disclosed, according to baseline variables*
Baseline variable Patient sex Female Male
No. 38 8
1.6 2.0 0.41
41 5
1.8 0.8 0.11
20 26
1.8 1.6 0.63
16 17 13
l.l 2.0 2.0 0.05t
17 16 13
1.1 2.2 1.8 0.05:1:
P
Patient race White Nonwhite p Patient education High school or less Beyond high school
Mean no. of side effects disclosed
P
Patient age
INFORMED CONSENT AND THE PRESCRIPTION OF NONSTEROIDAL ANTIINFLAMMATORY DRUGS JEFFREY N. KATZ, LAWREN H. DALTROY, TROYEN A. BRENNAN, and MATTHEW H. LIANG Objective. To examine disclosure of side effects of nonsteroidal antiinflammatory drugs (NSAIDs) and to identify patient- and physician-specific factors associated with greater disclosure. Methods. Forty-six encounters between rheumatologists and new adult outpatients who were prescribed an NSAID they had not been taking prior to the visit were audiotaped. Reviewers coded the NSAID prescribed, specific side effects mentioned, demographic features of patients and physicians, and patient clinical characteristics. Neither patients nor physicians were aware that side effect disclosure was being studied. Results. A mean of 1.7 side effects was mentioned From the Departments of Medicine and Rheumatology! Immunology, Brigham and Women's Hospital, the Robert B. Brigham Multipurpose Arthritis Center, and the Department of Health and Social Behavior, Harvard School of Public Health, Boston, Massachusetts. Supported by NIH grant AR-36308 to Drs. Katz, Daltroy, and Liang. Dr. Katz is recipient of a Postdoctoral Research Fellowship from the Arthritis Foundation. Jeffrey N. Katz, MD, MS: Department of Rheumatology! Immunology, Brigham and Women's Hospital, and the Robert B. Brigham Multipurpose Arthritis Center; Lawren H. Daltroy, DrPH, MPH: Department of Rheumatology!Immunology, Brigham and Women's Hospital, the Robert B. Brigham Multipurpose Arthritis Center, and Department of Health and Social Behavior, Harvard School of Public Health; Troyen A. Brennan, MD, JD, MPH: Department of Medicine, Brigham and Women's Hospital; Matthew H. Liang, MD, MPH: Departments of Medicine and Rheumatology! Immunology, Brigham and Women's Hospital, and the Robert B. Brigham Multipurpose Arthritis Center. Address reprint requests to Jeffrey N. Katz, MD, MS, Department of Rheumatology!Immunology, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115. Submitted for publication December 6, 1991; accepted in revised form April 30, 1992.
per encounter. Epigastric discomfort was mentioned in 72 % of encounters, while other side effects, including hepatic, renal, hematologic, or central nervous system effects, were mentioned in ::515% of encounters. Three factors were identified as independent predictors of less disclosure of side effects: senior clinician (versus less experienced), patient not taking another NSAID immediately prior to the visit, and patient age 7.0) for this analysis. Bleeding tendency was not rated by Fries et al and was designated in this study as a mild effect. Sixteen tapes were reanalyzed to explore the relationship between prior NSAID use and physician experi-
INFORMED CONSENT AND NSAIDs
1259
ence. Half the tapes were of encounters with senior clinicians half with less experienced physicians. For each encounter, reviewers coded whether side effects were first mentioned by the patient or by the physician. Statistical methods. Bivariate relationships between baseline features and the number of side effects disclosed during the encounter were evaluated using nonparametric (Wilcoxon rank sum) tests. Mult!pl~ linear r~gression (15), log-linear analysis (16), and logistic regressIo~ (17) were used to analyze the relationships between baseline featu~es and side effects in multivariate models. The number of SIde effects disclosed served as the continuous dependent variable in the linear regression. In logistic regression and log-linear analyses, the dichotomous outcome variable was whether 2 or more side effects were disclosed, versus 0 or 1. Categorical indicators for patient education level, sex, race, age, and diagnosis, and physician ca~eg?ry (see above) served as independent covariates. A significance level of 0.007 (0.05 divided by 7) was established to minimize the possibility of Type I error due to testing of 7 predictors.
Table 2. Disclosure of possible side effects of specific nonsteroidal antiinflammatory drugs
RESULTS
present analysis. Audiotapes were incomplete on 3 patients, leaving a study sample of 46 patients. Of 28 arthritis center physicians (79%) who were asked to participate in the study, 22 consented to do so. . Subject characteristics. The group of 46 patients was predominantly white and female (Table 1). The patients were divided evenly with respect to several study variables, including age category «40, 40-59, ~60), education level (some college versus none), experience of the physician they saw, disease category, and whether the patient was receiving a different NSAID or no NSAID treatment at the time of the visit. Ten different NSAIDs were prescribed in the 46 encounters, including naproxen (12 patients), piroxicam (8 patients), ibuprofen (7 patients), and diclofenac, tolmetin, meclofenamate, sulindac, indomethacin, aspirin, and salsalate (:54 patients each). The median number of encounters per physician was 3, with a range of 1-5. Disclosure. Disclosure about each of several specific adverse effects is presented in Table 2. While the possibility of epigastric discomfort was disclosed in 72% of encounters, none of the remaining side effects was mentioned in more than 15% of encounters. The total number of side effects mentioned in each encounter ranged from 0 to 5, with a median of 1.5 and a mean of 1.7. In 11 encounters (24%), no side effects were mentioned. The total number of side effects mentioned did not vary significantly across the different NSAIDs (Kruskal-Wallis ¥ = 15.0,9 degrees offreedom [dt], P = 0.09). There was little correlation (r = 0.10) between the number of side effects disclosed and patient satisfaction with the discussion of the therapy prescribed. The relationships between disclosure of specific side effects and their reported frequency and severity
Subject recruitment. In the parent study, we contacted 604 consecutive eligible patients who were planning a first visit to the clinic. Three hundred forty-five (57%) expressed interest in participating in a prospective study that involved audiotaping of2 visits, 2 telephone interviews, completing questionnaires on 3 instances, and random assignment to a counseling intervention or a control group. After exclusions for logistical reasons (scheduling conflicts and missed appointments), 178 patients were enrolled in the study. Of these, 134 were expected to continue their arthritis care at the clinic with the study physician, and data from their initial visits constituted the parent study sample. Forty-nine of these patients were prescribed a new NSAID at the initial office visit and therefore were eligible for the Table 1.
Baseline characteristics of the 46 patients studied Characteristic
Age 7.0) and 8% of possible less severe side effects, suggesting that disclosure of side effects was not influenced by their severity (I = 0.28, 1 df, P > 0.60). Bivariate associations between baseline variables and the number of side effects disclosed were analyzed with the Wilcoxon rank sum test and are presented in Table 3. More side effects were disclosed to patients >40 years old than to younger patients (mean 2.0 versus 1.1). Patient sex and education level were not significantly related to the number of side effects disclosed. More side effects were disclosed to white than to nonwhite patients (mean 1.8 versus 0.8) although, with only 5 nonwhite patients in the sample, the difference did not reach statistical significance (P = 0.11). Patients with either systemic inflammatory or chronic noninflammatory disorders had significantly more side effects disclosed to them than did patients with arthralgias or self-limited disorders (mean 1.8, 2.2, and 1.1 side effects, respectively). The strongest associations were observed for physician experience and whether NSAIDs had been used previously. Less experienced physicians disclosed more side effects than did senior clinicians (mean 2.1 versus 1.2) (although the difference [P = 0.02] did not reach statistical significance based on the stringent standard of P < 0.007 that had been established), and patients taking NSAIDs prior to the visit were told about significantly more side effects than were patients who were not taking NSAIDs prior to the visit (2.3 versus 1.1; P = 0.005). Several multivariate models were used to identify factors independently associated with the total number of side effects disclosed, controlling for all other variables. Stepwise linear regression confirmed that physician experience, patient prior use of NSAIDs, and patient age were associated independently with the extent of disclosure. Log-linear modeling with a bivariate outcome (0--1 side effects versus 2:2) also confirmed the independent predictive value of the same 3 variables. There were cells with 0 patients in the logistic regression and log-linear models, which produced unstable estimates of the model coefficients;
Table 3. Number of side effects disclosed, according to baseline variables*
Baseline variable Patient sex Female Male
No. 38 8
1.6 2.0 0.41
41 5
1.8 0.8 0.11
20 26
1.8 1.6 0.63
16 17 13
l.l 2.0 2.0 0.05t
17 16 13
1.1 2.2 1.8 0.05:1:
P
Patient race White Nonwhite p Patient education High school or less Beyond high school
Mean no. of side effects disclosed
P
Patient age
