Mechanisms Behind the Failure of Residents' Longitudinal Primary Care to Promote Disclosure and Discussion of Psychosocial Issues FREE

PRIMARY CARE is thought to create an atmosphere in which sensitive concerns are more readily raised.^1- 2 Studies in both pediatric and adult settings support this view—detection of distress is increased when physicians and patients (or parents) believe they have an ongoing relationship.^3- 5 Guidelines for pediatric primary care,⁶ supported by studies in child development,⁷ underline the importance of detecting problems with parental mental health, especially in the first years of a child's life. Studies looking at practices at a single point in time, however, have found low rates of identification of parental distress by resident and faculty pediatricians.^3,8 To our knowledge, no studies in pediatrics, and only a few in other specialties,^9- 12 have examined care over time to see if it leads to improved identification of mental health problems, or if there are specific primary care provider skills important to identification of distress in longitudinal settings.

Our goal was to examine the identification of parental distress in the course of longitudinal care in a pediatric residents' continuity clinic. Our specific aims were to (1) determine if the frequency of discussion, disclosure, and detection of parents' social and emotional concerns changed in the course of longitudinal care, and (2) describe physicians' communication behaviors that might be responsible. We focused on 2 physician communication behaviors. First, several studies have demonstrated that a cluster of behaviors, collectively known as patient-centeredness (ie, giving information, showing empathy, listening attentively, and asking questions about social and emotional issues) are associated with increased patient willingness to share concerns.^13- 15 Second, studies have found that physicians sometimes seem to ignore or underrespond to patients' tentative disclosures of distress.^16- 20 We hypothesized that patient-centeredness would be associated with increased rates of discussion and detection of parental distress and that avoidant responses would be associated with decreased discussion and detection.

The study was conducted in the pediatric primary care clinic of an urban teaching hospital. Children have assigned primary care providers whom they see for health maintenance and, when possible, acute care.

The study design was descriptive, using a prospective longitudinal design. Data were collected as part of a clinical trial (the SAFE Home Project) designed to test the effect of anticipatory guidance on parents' injury prevention practices. Physicians participating in the SAFE Home Project were randomized to be trained in injury counseling (two 2½-hour sessions with role playing and demonstrations) or to receive a single 1-hour seminar on injury prevention.²¹

Forty-four first- and second-year residents were eligible to participate in the SAFE Home Project; 31 residents (4 Asians and 27 whites) (70%) agreed to participate. Twenty-three (74%) were women and 15 (48%) were in their first-year of residency. Parents (or other guardians) were eligible if their child was 6 months of age or younger and cared for by a study-enrolled physician. Parents were ineligible if they did not speak English or did not live with their child. Research assistants systematically approached 224 eligible parents over an 8-month period; 196 (88%) agreed to participate; 25 (11%) declined, and 3 were ineligible. Of the 196 enrolled, 190 (85% of those approached) had at least 1 recorded visit and were included in our analysis. Parents and primary care providers gave informed, written consent for participation. The study was approved by the Joint Committee on Clinical Investigation of Johns Hopkins Hospital, Baltimore, Md.

One hundred eighty-three parents (96%) were the child's biological mother; 3 were female guardians, and 2 were fathers. Parents' average age was 24 years (median age, 22 years; age range, 15-64 years) and infants were, on average, 2½ months old at the time of enrollment (median age, 2 months; age range, 3 days to 6 months). Ninety percent of the parents were African American, 8% were white. One was Hispanic and 2 were Native American. Fifty-one percent graduated from high school and 14% had taken some college courses.

We calculated Bice and Boxerman's²² Continuity of Care (COC) index using data from the outpatient appointment system.²³ The COC index differs from a simple percentage of visits with one's physician in that it is also sensitive to the total number of physicians seen. Values range from 0 to 1, where 1 indicates all care provided by a single physician. From birth to the end of the study, enrolled infants and parents made 1538 visits to the primary care clinic, 94 to the emergency department, and 276 to other hospital clinics. On average, infants saw their assigned physician for 92% of the primary care visits (range, 22%-100%) and 71% of all visits at the medical center (range, 6%-100%). The COC index averaged 0.72 (range, 0-1) and was not significantly related to the physician's sex (male, 0.71; female, 0.73; P = .78).

Research assistants attempted to audiotape consecutive visits between enrolled parents and physicians. Six hundred ninety-two visits were recorded, a median of 4 per parent-physician pair (range, 1-9) and 21 per physician (range, 4-48). For individual parents, recorded visits spanned an average of 27 weeks from first to last visit (range, a single visit to 57 weeks). Of first-recorded visits, 50% were of the parent's actual first or second visit with the physician, while the remainder of first-recorded visits ranged from the 3rd to the 17th actual visit. Most (82%) of the recorded visits were for well-child care. Visits lasted an average of 25 minutes (median, 25 minutes; range, 2-91 minutes).

Parent-physician talk was coded using the Roter Interactional Analysis System (RIAS).¹⁵ The Roter Interactional Analysis System coders listen to audiotapes and classify each speaker's utterances into one of several categories including information giving, question asking, empathy, and partnership facilitation. Reliability is monitored by duplicate coding of a random 10% sample of audiotapes. Correlation coefficients for individual Roter Interactional Analysis System categories ranged from 0.75 to 0.97. Physicians' patient-centeredness was calculated first as the sum of physician talk in information giving (medical and psychosocial), asking psychosocial questions, showing empathy, giving reassurance, and partnership building.²⁴ Some formulations of patient-centeredness do not include medical information giving,²⁵ so we calculated a second sum, partnership building omitting medical information giving. The 2 measures were highly correlated (r = 0.96). The Roter Interactional Analysis System coders also noted if visits included 3 categories of social and emotional topics (parental mood, social and family issues, and referrals for mental health care), and who initiated the topic (parent or physician). Discussions of these topics were transcribed and coded for the physician's method of initiation (neutral or positive-leading tone) and response. Physician responses were classified using Hadjiisky and colleagues'¹⁶ typology of avoidant or discouraging clinician responses (Table 1).

At each recorded visit, parents completed the 28-item General Health Questionnaire (GHQ), a well-validated brief screen for adult emotional distress.^8,26- 28 At a cutoff score of greater than 4 (used in this study), sensitivity for psychiatric disorder is more than 85% with a specificity of more than 75%. Forty-eight (25%) of the 190 parents scored positive on the GHQ at 1 or more visits (n = 68). Visits where the parent scored positive on the GHQ were distributed evenly across the time span of parent-physician relationships. After each visit, physicians rated the parents' emotional health on a scale of excellent, good, fair, or poor.

The multiple visits of each parent-physician pair, and the visits of multiple parents with the same physician, are not statistically independent. After exploratory analyses using standard tests (χ² and t tests, ordinary linear and logistic regression), we computed final results using generalized estimating equation–based procedures in the statistical software Stata Version 6.0 (Stata Corp, College Station, Tex). These procedures compute population-averaged statistics that consider the nonindependence of observations, and also require few assumptions about the distributions of the variables explored.²⁹

A second problem involves missing observations. Overall, we had recordings for 692 (68%) of 1015 visits between parents and their infant's primary care physician from the time they enrolled in the study to their last recorded visit. The generalized estimating equation–based methods described above provide good estimates when data are "missing at random," that is, when whatever mechanism accounts for the pattern of missing data does not depend on the outcome being studied.²⁹ To test if our data met this condition, we developed regression models to predict if the next visit would be recorded, based on whether a parent or physician initiated psychosocial discussion at the immediately prior visit. These models did not yield statistically significant coefficients.

Social and/or emotional topics were mentioned during 231 (33%) of the 692 recorded visits (Table 2). Social and family issues were the most frequent (30% of visits). Parent's sleep habits or mood was mentioned in 8% of the visits, and mental health referrals or treatment were mentioned in 3%. More visits contained physician-initiated topics (27%) than parent-initiated topics (8%).

Visits with female physicians had higher patient-centeredness scores (average, 259 utterances per visit for female vs 198 for male; mean difference, 60.6; 95% confidence limits [CLs] for difference, 38.5%, 82.6%; P<.001). Assignment to the SAFE Home Project intervention group was associated with a significant increase in patient-centeredness for female but not male physicians' visits (average increase for female physicians, 31.7%; 95% CLs, 10.6%, 52.7%; P = .003; average increase for male physicians, 18.2; 95% CLs, −17.4%, 53.8%; P = .31).

Patient-centeredness scores were significantly higher in visits where social or emotional issues were mentioned, compared with visits where parents did not initiate these discussions (Table 3). In visits in which parents initiated discussion of social or family topics, physicians made on average 100 more patient-centered utterances (95% CLs, 65.7%, 133.9%, about a 40% increase).

Parental distress increased the odds that parents, but not physicians, would initiate discussion of psychosocial topics. Visits where a parent was distressed were more likely to contain a parent-raised social or emotional topic (odds ratio, 2.4; 95% CLs, 1.1%, 5.1%; P = .03, adjusted for patient-centeredness, visit number, COC index, physician's sex, and intervention group) (Table 4). Visits with distressed parents were not more likely to contain discussion of a physician-raised topic (odds ratio, 1.3; 95% CLs, 0.73%, 2.2%; P = .42).

The odds that parents would initiate discussion of social or emotional topics did not change significantly over the course of time (Table 4). The odds that a parent-initiated topic would be discussed at visits 6 or more vs during visits 1 through 5 was 0.84, with a 95% confidence interval of 0.44 to 1.6. Similarly, distressed (GHQ-positive) parents were no more likely to initiate discussions at visits 6 or more compared with visits 1 through 5.

Time tended to decrease the odds that physicians would raise social or emotional issues, even after adjustment for patient-centeredness that, as noted earlier, is positively associated with the physicians' initiation of these topics. The odds of physicians raising topics at visits 6 or more, vs visits 1 through 5, were 0.46 (95% CLs, 0.31%, 0.67%). The COC index was not associated with an increased odds of physician-initiated discussions.

Among 587 visits where both parent GHQ and physician ratings were available, physicians' sensitivity was 14% (9/63) and specificity 90% (472/524). Accuracy (correctly rating the parent as distressed or not distressed) was unrelated to patient-centeredness, sex, intervention group, continuity, or the number of visits (Table 5).

Of 141 social and/or emotional discussions initiated by physicians in transcribed segments, 37 (26%) began with a leading positive question or statement. For example:

Physician: Sounds like you really enjoy being with her. Mother: Mm. Physician: You look so happy when you're with her. Mother: (Laughs) She's a good baby.

Compared with more neutral statements or questions ("What has it been like having a new baby?"), positive openings were less likely to be followed by parents' disclosure of a problem. Of 37 discussions initiated with a positive question or statement from the physician, 7 (19%) resulted in disclosure of a problem compared with 52 (50%) of 104 episodes (odds of disclosure given positive opening, 0.23; 95%CL, 0.08%, 0.62%; P = .002).

In 116 (53%) of 220 transcribed social and/or emotional discussions, parents indicated a possible or definite problem, or asked for help. In 47 discussions (41%), physicians followed up in ways that seemed to avoid or discourage further talk (Table 1). Confused or powerless responses were the most common (15/116 [31%]), followed by ignoring or changing the subject (15/116 [13%]), overly authoritative responses (12/116 [10%]), and nonhelpful reframing (5/116 [4%]). These responses were associated with a decreased likelihood of parents disclosing a problem in a subsequent discussion in the same visit or during the next recorded visit. To perform this analysis, we first identified all discussions of psychosocial problems (n = 45) for which there was either a second psychosocial discussion in the same visit or for which there was at least one subsequent recorded visit. Of 25 problem discussions with avoidant responses, 6 (24%) were followed by disclosure of a problem. Of 20 problem discussions without an avoidant response, 11 (55%) were followed by disclosure of a problem (odds ratio, 0.26; 95% CLs, 0.07%, 0.90%; P = .03).

In this pediatric residents' clinic, only 15% of visits by distressed parents included a parent-initiated discussion of social or emotional topics. There was no increase over time in the odds that psychosocial topics would be raised by parents or physicians (regardless of parental distress), and time did not improve the odds that physicians would accurately identify a parent's emotional state. These observations are consistent with the few longitudinal descriptions of psychosocial discussions in adult primary care.^9- 12

Patient-centeredness, avoidant responses, and positive leading questions partly explain these time-related trends. Physicians' patient-centered talk was a strong predictor of both physician- and parent-initiated discussions of social and/or emotional topics, and it tends to decrease over time.³⁰ Physicians' avoidant responses—made in response to about 40% of parent problem statements—were associated with a decreased probability of disclosure. Levinson et al²⁰ found that community-based internists and family physicians responded positively to only about 20% of clues to emotional distress offered by their patients. Other studies in pediatrics have found similar low rates of response and low levels of empathy.^17- 18,31 We also found that questions asked with a positive bias resulted in fewer disclosures of social and/or emotional problems. It is possible that residents may have felt that these young, relatively disadvantaged mothers needed extra encouragement, or that being overly enthusiastic reflected compensation for so frequently feeling unable to help in other ways.

To our knowledge, this is the first large-scale analysis of the evolution of communication in primary care. Although our findings are consistent with observations from other adult and pediatric studies, several aspects of the population that we studied might limit generalizability. We studied residents, but studies of community physicians caring for adults have shown similar trends,^9- 11 and a study of 10 private practice pediatricians reported that only 62% of the parents felt the physician had listened to their ideas.³² Residents may also have different attitudes toward involvement with patients, knowing that there is a time limit on the length of their relationship. However, the level of continuity achieved in the clinic we studied (overall COC index of 0.72) is comparable to figures reported from 2 private pediatric practices and higher than figures from another teaching hospital's resident clinic.³³ Finally, we studied pediatricians' discussions of parental rather than child problems. Pediatric health supervision guidelines place a heavy emphasis on elicitation and discussion of parental mood, stress, and family dynamics,⁶ but we do not know how those guidelines are regarded by pediatricians in general or by the residents we studied in particular.

Another limit to generalizability involves the study's parents, who were largely young African Americans and from low-income neighborhoods. In a diverse group of US practices, depressed African American adults were less likely than whites to have their condition detected.³⁴ Another study in adult primary care found that African American patients perceived white physicians as having a less participatory interactional style compared with African American physicians.³⁵ There were no African American physicians in our study. Detection may also have been reduced because parents were not sufficiently distressed. Greater severity of distress increases the odds of physicians detecting adult emotional distress.^3,19,34 In our study, among parents who screened positive on the GHQ, the average score was 8.6 (range, 5-21), in a low-moderate range.³⁶ However, our residents' sensitivity to distress (14%) is similar to that reported for residents by Heneghan et al³(11%).

Another limitation is that 74% of the physicians in our study were women, somewhat higher than the proportion of women in categorical pediatric residencies in the United States, which has been slowly increasing among first-year resident classes from 60% in 1991 to 66% in 2000.³⁷ Female physicians tend to have longer visits with their patients and engage in more counseling, partnership building, and listening.³⁸ One explanation advanced for these differences is that women may believe more than men in the healing power of relationships, show less tolerance of detachment, and are less willing to trade intimacy for achievement.³⁹ In our study, female physicians were more patient-centered, and when trained in injury counseling had a greater increase in patient-centeredness, but were no more likely than male physicians to correctly identify parents' emotional states.

We also had to account for difficulties with data collection. We compensated by analyzing data according to each visit's real position in the parent-physician sequence, and by using statistical techniques designed to accommodate missing data when calculating average trends over time. We were able to satisfy ourselves that visits were missing randomly with respect to our main outcome measures, but it is still possible that our analysis of avoidant responses missed key visits at which important topics were discussed.

Our results suggest that creating opportunities for longitudinal relationships does not itself lead to increased discussion and detection of parental psychosocial issues. The results do not imply that longitudinal or high-continuity care is no better than episodic or low-continuity care; we did not make comparisons with parents whose children lacked a designated primary care provider. We do propose possible explanations for why mental health detection rates remain low, even in continuity settings.^{3,34,36,40- 41} Three core sets of skills—actively maintaining a patient-centered orientation, learning how to respond to concerns, and being encouraging but neutral when asking questions—have the potential to improve disclosure of distress during longitudinal care. These skills can be taught to physicians at various stages in their careers.^42- 44 These skills, however, may be necessary but insufficient to improve mental health outcomes in primary care. Patient-centeredness was associated with more psychosocial discussion but not increased detection. Residents likely also need training in how mental health problems present and their treatment.^45- 46

Accepted for publication March 29, 2002.

This study of the analysis of longitudinal care was supported by grant MH-57782 from the National Institute of Mental Health, Bethesda, Md. The SAFE Home Project study was supported by grant MCJ-240638 from the Maternal and Child Health Bureau, Health Resources and Services Administration, Rockville, Md.

This study was presented in part at the National Institute of Mental Health conference, Challenges for the 21st Century: Mental Health Services Research, Washington, DC, July 19, 2000, and at the International Conference on Communication and Health, Barcelona, Spain, September 21, 2000.

Longitudinality and continuity are believed to facilitate disclosure of emotional distress and other sensitive issues. Studies in both adult and child primary care, however, find low detection rates for distress, and reports that physicians frequently ignore or respond minimally to patient concerns. No studies to date have examined sequences of primary care visits to determine mechanisms that might promote or discourage detection of distress.

This study found that in a pediatric residents' continuity clinic, neither longitudinality nor continuity appeared to promote discussion or detection of parental distress. Three core sets of communication skills—actively maintaining a patient-centered orientation, response to concerns, and the manner of asking questions—were associated with changes in the odds of parents' discussion of distress. Simply providing continuity experiences may not be sufficient to train residents in primary care; attention may also be needed to specific interpersonal skills needed to promote relationships over time.

Corresponding author and reprints: Lawrence S. Wissow, MD, MPH, Bloomberg School of Public Health, Johns Hopkins University, 624 N Broadway, 749 Hampton House, Baltimore, MD 21205 (e-mail: Lwissow@jhsph.edu).