3.1 Patient and visit characteristics
Of the 58 patients, the mean age was 42.7 years (SD 7.5), 31% were female, 49.1% were white, 26.3% African American, 21.1% Hispanic, and 3.5% were of other race/ethnicity; 55.2% were heterosexual, and the median CD4 count was 314 cells/mm3. Having detectable viral load was a selection criterion, but most were not very high (median 432 particles/mm3, 25th percentile 75, 75th percentile 5,731). Observed visit length was quite consistent with previous studies, with a mean duration of 16:31 minutes and a median of 15:27 minutes. (Clock time did not include periods when the physician was not in the room.) The median utterance count was 539, ranging from 130 to 1,490. shows the distribution of the various measures.
Distribution of patient-centered care measures and visit length
3.2 Bivariate relationships
All five measures of patient centeredness were significantly positively associated (p<.05) with visit length, with rank order correlations ranging from .21 to .44 after adjusting for clustering within patients (). These parameters can theoretically range from 0 to 1. They correspond to the change in rank of the dependent variable associated with one change in rank of the independent variable. The pattern of relationships between indicators of patient-centeredness and clock time was similar (data not shown.)
Rank order correlations for measures of patient centeredness.*
3.3 Multivariable relationships
shows two sets of models. Model 2 is adjusted for race, study arm (control vs. intervention), and site (parameters not shown), which were the only statistically significant covariates. Only one of the sites, a group practice, was significantly different from the reference site in visit length, tending to have longer visits. In four cases, the adjustment attenuated the relationship between the measure of patient centeredness and visit length. For “percent of patient expressives” and “ratio of physician to patient utterances” the adjustment caused the p-value to exceed .05. For “percent patient control utterances” the parameter estimate increased with adjustment, and the p-value changed from 0.0044 to 0.0009. The negative binomial coefficients refer to log ratios. A coefficient of 0.1 is interpreted as follows: if the indicator of patient centeredness were increased by 10%, the visit would be exp(0.1) = 1.105 times as long.
Bivariate and multivariable models of visit length
Patient race/ethnicity and the intervention indicator were significantly related to visit length in all adjusted models. Black race/ethnicity was significantly associated (p<0.05) with shorter visits in all models, and Hispanic race/ethnicity in three of five models. Being in the control arm of the experiment was also consistently associated with shorter visits. This was not surprising as the purpose of the intervention was to increase discussion about antiretroviral medication adherence.(14
3.4 Qualitative description
The topics are largely driven by the patients in four of the five longest visits. A similar proportion holds in the sixth through fifteenth longest visits. Two are physician dominated and driven by medical complexity, in six the patient does much of the agenda setting, and two of them are relatively balanced in the proportion of patient- and physician-initiated topics. In the following, we number cases by their rank in visit length.
Case 2 is the same physician-patient dyad as Case 12. In both visits the patient initiates a large amount of discussion about issues which we coded as Logistics, such as not liking the way she is treated by clerical and nursing personnel, wanting to have her ob/gyn care transferred to another institution, and wanting office-based methadone treatment rather than attending a methadone clinic. There is also considerable patient-driven biomedical content which is not of great concern to the physician, such as skin tags and spider veins, and both visits feature a great deal that we coded as Socializing.
One way of looking at this is that so much of this patient’s time and personal relationships are bound up with her medical care that the distinctions among logistics, biomedical, psychosocial and socializing are not salient for her. On the other hand these interactions also include what might be characterized as digressions, and raise boundary issues. For example, the physician spends considerable time teaching the patient about aspects of recent world history, discusses personal information about another of her patients, and shares an intimate confidence about her family with the patient.
Case 1 also raises boundary issues. The physician is female, and the patient is a gay man who gave up his pattern of casual sex with multiple partners, and drug abuse, after learning that he was HIV positive. He has abandoned his former friendships and is now socially isolated. The physician initiates and pursues a long discussion of this common problem. The physician finally expresses personal fondness for the patient, and asks to meet in a social setting. The patient seems uncomfortable with this suggestion.
Case 3, a woman in recovery, is also struggling to stay away from old associates, once again be a good mother, and rebuild a constructive social life. The physician is consistently affirming, positively reinforces the patient’s goals, and listens reflectively.
Case 4 features predominantly biomedical content but it is driven by concerns raised by the patient, who works as an allied health professional. There are long stretches consisting of the physician providing information, but in response to explicit patient requests. For example, the patient wants liposuction because of liposdystrophy, and is transgendered. The physician spends considerable time discussing whether various plastic surgeons will constitute a good referral for a transgendered person. The visit also features much socializing, including banter about a popular singer, and the physician criticizing colleagues who work in the same practice about matters unrelated to the patient’s care.
Case 5 is long because of biomedical complexity. The patient has viral drug resistance and is pregnant, and the physician is considering changing the regimen. This physician is highly directive and spends some time “lecturing” the patient about the importance of strict adherence. The physician’s agenda clearly dominates in this visit, and there is little psychosocial content.
Cases 14 and 15 are the same physician-patient dyad. In case 14, the patient tells a long, emotional story about his mother’s recent death and squabbling with his siblings over the estate. In the second visit, the patient comes in with a lengthy written list of symptoms and biomedical concerns. In both cases, the patient’s agenda dominates, but one is mostly psychosocial, the other mostly biomedical.
There is less to be said about the 15 shortest visits because, obviously, there is much less content. Although some are coded as having an above-average proportion of psychosocial content, this is never very meaningful. For example, in one case the physician merely asked if the patient was still using heroin, and the patient responded negatively. Since the visit was so short, this still registered as a substantial percentage of the total. For the most part, these visits are highly biomedically focused. Patients ask few or no questions, and do not initiate topics. Patient responses to physician questions are brief. The visits largely consist of review of medications, review of systems, and a brief inquiry about adherence.
In two of the short visits, there are special circumstances. In one, the physician thought the patient had come in only to have a wart removed. The physician confessed that he had not reviewed the patient’s lab reports. After a brief social chat, they rescheduled. In another case, the patient was in a hurry to get to a temporary job, and grew anxious as the physician extended the visit.