Ming Tai-Seale, PhD

Periodic health examinations (PHEs) are the most common reason adults see primary care providers. It is unknown if PHEs serve as a “safe portal” for patients with mental health needs to initiate care. We examined how physician communication styles impact mental health service delivery in PHEs.

Retrospective observational study using audio-recordings of 255 PHEs with patients likely to need mental health care.

Mixed-methods examined the timing of a mental health discussion (MHD), its quality, and the relationship between MHD quality and physician practice styles. MHD quality was measured against evidence-based practices as a 3-level variable (evidence-based, perfunctory, or absent). Physician practice styles were measured by: visit length, verbal dominance, and elicitation of a patient’s agenda. A generalized ordered logit model was used.

 Many patients came with mental health concerns, as over 50% of the MHDs occurred in the first 5 minutes of the visit. One-third of the 255 patients had an evidence-based MHD, another third had a perfunctory MHD, and the remaining had no MHD. MHD quality was significantly associated with physician communication styles. Visits with physicians who tend to spend more time with patients, fully elicit patients’ agendas, and let patients talk (instead of being verbally dominant) were more likely to deliver evidence-based MHD.

If done well, PHEs could be a safe portal for patients to seek mental health care, but most PHEs fell short. Improving PHE quality may require reimbursement for longer visits and coaching for physicians to more fully elicit patients’ agendas and to listen more attentively. 

Periodic health exams (PHEs) could be a “safe portal” for patients with mental health needs to receive care. We found:

The quality of mental health discussions varied greatly in 255 audio-recorded PHEs in an 

integrated delivery organization: 1 of 3 was evidence-based, 1 of 3 was perfunctory, and 1 of 3 was nonexistent. 

Physicians who spend more time with patients, fully elicit patients’ agendas, and let patients talk were more likely to deliver evidence-based mental health care. 

Improving care quality may require reimbursement for longer visits, coaching for physicians to fully elicit patients’ agendas, and to listen more attentively. 

Routine assessment of mental health status should be reinforced.

The majority of patients with mental health concerns turn to their primary care physicians for help.

 Many rely on periodic health examinations (PHEs)—the most common reason for adults to visit physicians.

 Proponents of PHEs argue that, besides increasing the use of preventive services, PHEs bring patients into contact with their primary care physicians, and therefore, provide a route by which patients seek services that they feel uncomfortable disclosing to the scheduling staff. In fact, as many as 1 in 3 patients have undisclosed reasons for requesting a PHE.

 Mental health care needs may be an example of undisclosed reasons that prompt some patients to use the time-honored PHE tradition as a “safe portal” to seek mental health care.

Evidence-based mental health care improves patient outcomes not only through the reduction of mental illness symptoms, but also by improving physical and social functioning.

 Primary care physicians’ communication practices and skills have been documented to influence the quality of mental health service delivery in primary care visits.

 It has been shown that even when patients have an agenda, they commonly do not make it explicit in ambulatory encounters.

 Essential communication approaches that can be employed by the physician to ensure effective communication (eg, being open to patients’ agendas

), fully eliciting patients’ concerns and preferences,

 and co-creating the visit agenda with them,

 contribute to improving the overall quality of patient—provider communication.

In addition to fully eliciting patients’ agendas for the visit, physicians must give patients enough time to speak. It is not uncommon that some physicians dominate the conversation, in order to control the time used; however, it is necessary to balance the need to manage the conversation with the need to let the patient speak. Verbally dominant physicians disempower patients and impair patient engagement.

 It has been well-documented that patients were less satisfied with their physicians when physicians talked more and when patients perceived their physicians as domineering, and patients were ultimately less likely to sue physicians with low verbal dominance.

 The literature is relatively silent on how primary care physicians’ verbal dominance could affect the quality of discussions about mental health during PHEs.

Giving patients time to speak rather than dominating the conversation can affect the length of visits, however. Visit length has long been the subject of research and has been shown to be associated with the delivery of evidence-based preventive health services (both screening and counseling).

 It is not known if patients with mental health needs who see primary care providers (PCPs) who tend to provide longer visits with patients may be more likely to receive evidence-based mental health services. Mental health issues are perceived to be difficult and have been well-documented to receive suboptimal attention in ambulatory care settings.

 Furthermore, the provision of mental health screening and counseling during PHEs has received less attention than the delivery of evidence-based biomedical screening and lifestyle counseling during PHEs.

For PHEs to be valuable to patients with mental health needs, evidence-based mental health discussion is required. Figure 1 illustrates our hypothesis that 3 sets of predisposing physician and patient factors influence the quality of mental health discussion (MHD): evidence-based, perfunctory, or absent. The first group of factors consists of physician practice styles: how much time a physician usually provides in visits, a physician’s openness to the patient’s agenda, and a physician’s verbal dominance. The second set of factors relates to patients’ mental health status and if the patient is in an ongoing episode of care

 for a mental health issue. The third factor relates to a patient’s preparedness for the visit (ie, if a list has been prepared and brought to the visit for discussion with the physician). These tasks have also been shown to increase patient satisfaction and subsequent patient self-management.

Using qualitative and quantitative research methods, we investigated whether PHEs offer patients an opportunity to discuss mental health with physicians. We also measured the quality of MHD based on its concordance with evidence-based practices by coding audio recordings and transcripts of PHEs (described later). We further sought to operationalize measures of physician communication styles that could be associated with the delivery of evidence-based MHD and are also potentially malleable.

Patients with mental health needs were drawn from a sample of 484 participants, aged 50 years or older, in a study of preventive health discussions,

 which took place at an integrated health delivery system in Detroit, Michigan, between February 2007 and June 2009. Patients completed a brief telephone survey at recruitment, containing the Patient Health Questionnaire-2 (PHQ-2)

 and sociodemographic characteristics. Furthermore, each patient’s visit was observed and audio recorded by a research assistant. Additional details of the study have been reported elsewhere.

 The institutional review boards of relevant organizations approved the study. Informed consent of study participants conveyed that the study would examine patient—physician communication about preventive health issues. No specific mention of mental health focus was made.

The sample for the current study consisted of 56 PCPs and 255 patients identified as likely in need of mental health care if they met any of the following criteria: 1) scored ≥2 on the PHQ-2,

 2) filled or were prescribed a psychotropic medication, 3) had a mental health diagnosis (

International Classification of Diseases (ICD)-9

 codes 290, 293-302, 306-316) in the electronic health record (EHR), or 4) visited a behavioral health provider. The look back period was 12 months before the visit.

 Five researchers coded audio recordings and transcripts of the visits to capture topics within 7 major areas: biomedical (eg, high blood pressure), health behaviors (eg, smoking), mental health (eg, depression,), psychosocial (eg, family), physician—patient relationship (eg, physician availability), visit flow management (eg, agenda setting), and other (eg, small talk about weather). Topics were defined as issues that had at least 2 complete exchanges between patient and physician. The time spent on discussing each topic was also recorded for both the patient and the physician. This analytical approach has been described in detail and applied in previous research.

 Studying a visit as a conversational event enabled us to understand the relative time spent on each topic by the patient and the physician, and if and how one person dominated the conversation.

Scores from different raters were compared using intra-class correlations for numerical variables and percentage agreement for categorical variables. Intra-class correlations between raters and within the same rater ranged from 0.78 to 0.99.

The dependent variable was a 3-level variable for the quality of MHD. We defined MHD as any exchange about depression, general anxieties and worries, emotional distress, death, bereavement, grief, mourning, death of others, pain, suffering, concerns, and worries regarding one’s own physical condition, tests, treatments, procedures, or other mood disorders.

 Whether or not the MHD was evidence-based was determined by the degree of concordance with treatment guidelines,

 including if the physician assessed the patient’s mood using any item from the PHQ-9,

 made a mental health diagnosis, prescribed psychotropic medication, made a referral to a mental health specialist, or made a plan for active surveillance of mental health symptoms (Figure 1). Perfunctory discussions (eg, “Any anxiety or depression?”) followed immediately by a nonrelated statement or questions (eg, “Any vaginal spotting or bleeding?”) were coded as non—evidence-based. The value of the outcome measure is 0 for no MHD, 1 for a perfunctory MHD, and 2 for an evidence-based MHD.

Key explanatory variables included both measures of physician practice style and patient characteristics. Physician practice style measures included visit length, eliciting patient agendas, and verbal dominance gathered from their visits with other study patients to form exogenous measures of these constructs to the index visit. Visit length was measured by the face-to-face interaction time in minutes between patients and physicians from other visits. Eliciting patient agenda was defined as the proportion of visits in which the physician attempted to fully identify the patient’s agenda (5 on a scale of 1 to 5)

—there was unanimous agreement among the 5 coders for this variable. Verbal dominance was defined by the ratio of actual talk time by the physician divided by talk time by the patient. To account for physicians’ proclivity to provide evidence-based services, we also included a count of evidence-based preventive services.

Patient characteristics were obtained from the EHR, a pre-visit patient survey, and direct office visit observation. Patient health status included whether in an ongoing episode of care (EOC) for mental illness,

 PHQ-2 score, and if the patient had brought a list of issues to discuss.

For the quantitative analysis, we specified a generalized ordered logit model with partial proportional odds for the 3-level ordinal dependent variable of evidence-based MHD. Our model constrained the odds ratios to be proportional across these 3 levels for all variables except PHQ-2 score. Tests of the proportional odds assumption indicated it was reasonable for the remaining variables. Three groups of explanatory variables were included in the model, as illustrated in Figure 1. The first group included physician practice style factors, as measured in the other visits to the same physician among the study sample: a) average visit length, b) percent of visits in which they fully elicited the patient’s agenda, and c) verbal dominance.

The second group were patient factors: patient’s mental health needs (ie, self-reported depressive symptoms in the PHQ-2),

 anxiety (ie, self-reported anxiety attack in the previous 4 weeks), and whether the patient was in an ongoing episode of care for mental health.

 The third related to patient activation level (ie, whether the patient brought a list of issues to discuss with the physician). Finally, we controlled for patient demographics (ie, age, sex, race/ethnicity, and education), comorbidity (Charlson Comorbidity Index score

), and the number of evidence-based services delivered in the visit.

 Standard errors were clustered by physician. The statistical analyses were conducted using Stata 14 (STATA, College Station, Texas).

To provide some contextual information on how visits between physicians at different levels of verbal dominance might evolve differently, we selected a few visits with high versus low verbally dominant physicians seeing patients with similar PHQ-2 scores to examine the timing and quality of mental health discussion. We mapped the topics, sequentially, as they took place during the visit.

Visit, Patient, and Physician Characteristics

Table 1 summarizes the characteristics of the study sample: 255 patients likely in need of mental health care saw 53 physicians practicing in about 2 dozen primary care clinics. The majority of patients were white (66%) and female (63%), with an average age of 60. About 9% of patients reported having an anxiety attack in the 4 weeks prior to study recruitment, the average PHQ-2 score was 1.1 [standard deviation (SD) = 1.5], and 37% were in an ongoing episode of care for a mental health condition. Their Charlson Comorbidity Index scores averaged 0.8 (SD = 1.4); 11% of patients brought a written list. These did not vary by MHD levels.

The median visit length was 26 minutes (SD = 10.3). A third of visits (33%) contained no MHD, 33% had an evidence-based MHD, and 34% had a perfunctory MHD.

Among the physician factors, the average length of other nonindex visits was 27 minutes (SD = 8) and rose with the level of MHD, from absent to perfunctory to evidence-based: 25 (SD = 6), 28 (SD = 9), and 29 (SD = 8), respectively. The average percent of visits in which the physician fully elicited the patient’s agenda in other nonindex visits was 25% and increased with the level of MHD, from absent to perfunctory to evidence-based: 19, 27, and 29, respectively. The average physician verbal dominance in other visits was 3.2 (SD = 2.2) and declined in visits with absent, perfunctory, and evidence-based MHD: 3.8 (SD = 2.8), 3.2 (SD = 2.1), and 2.6 (SD = 1.2), respectively. Finally, the average number of evidence-based services delivered was 2.9 (SD = 1.5) and increased with the level of MHD from absent to perfunctory to evidence-based: 2.7 (SD = 1.4), 2.8 (SD = 1.4), and 3.2 (SD = 1.6), respectively.

Mental Health as a Reason for Some Patients to Schedule a PHE

Over 50% of MHDs occurred within the first 5 minutes of visit initiation (median = 4.9; SD = 9.1). Ninety percent of them occurred within the first 19 minutes of the visit, which is 7 minutes fewer than the median visit length. The median time of MHD initiation was 6.3 minutes for perfunctory versus 3.3 minutes for evidence-based discussions (

 <.05). Thus, evidence-based MHDs occurred 3 minutes earlier in the visit than perfunctory MHDs.

In addition to MHDs occurring in the first few minutes of the visit, the nature of the conversation also suggested that mental health might have been a reason that some patients scheduled a visit. For example, one patient said that she “might need kind of an antidepressant” 54 seconds into her visit. She then broke down in tears and told the doctor that her sister was recently diagnosed with lung cancer. Another patient started to cry before the physician closed the exam room door, saying, “And I knew as soon as I saw you I would start to cry.” She stated that she was stressed out at work, “wake[s] up in the middle of the night, worry, worry, worry,” and then said, at 1 minute 24 seconds, “And I want a happy pill. Is there such a thing?”

Factors Associated With the Quality of Mental Health Discussion

Figure 2 shows results from the proportional logistic model for the 3 levels of MHD. The model cumulates over adjacent levels of the 3-level MHD quality outcome variable to form odds ratios. Our model assumes the odds ratios are the same for evidence-based MHD versus the combination of perfunctory MHD and no MHD, as they are for the combination of evidence-based MHD and perfunctory MHD versus no MHD. For ease of exposition, we state that each odds ratio is the proportional change in the odds of a “higher-quality visit” for each unit increase in the explanatory variable (scaled to SDs for the continuous variables: age, Charlson Comorbidity Index score, physician verbal dominance, visit length, and PHQ-2 score). For every SD increase in length of the physician’s other visits (SD = 7.8), the odds of having a higher-quality visit was 1.4 times greater. Similarly, the odds of having a higher-quality visit were 2.7 times higher among physicians who fully identified their patients' agenda (5 on the original 1 to 5 scale) compared with any other level of agenda setting. Increased physician verbal dominance is associated with lower odds of a higher-quality visit. Female patients were twice as likely as males to receive a higher-quality visit. White patients’ odds were 1.8 times higher than nonwhites’ of receiving higher-quality visit. Higher PHQ-2 scores (SD = 1.5) were associated with 1.8 times higher odds of evidence-based MHD compared with the combined perfunctory and no MHD, but not with increased odds of having any MHD versus no MHD.

Mental Health Discussion and Physician Verbal Dominance

Below we describe 2 mental health discussions (one perfunctory, the other evidence-based) that exemplified physicians with different levels of verbal dominance.

Patient 1’s PHQ-2 score was 4, indicating a high likelihood of depression. The patient had no other mental health diagnoses besides hyperkinetic syndrome in her childhood, recorded in the EHR. Physician A’s verbal dominance score for other sample visits was 9.7 (almost 3 SDs above the mean), indicating a tendency for Physician A to have a more dominant communication style relative to other physicians in the study. (There were only 5 physicians whose verbal dominance score was 10 or higher.)

Figure 3 illustrates the conversation flow between Physician A and Patient 1. Each color represents the talk time spent by the participant for each instance: dark blue for patient and light blue for physician. The longest biomedical exchange contained topics encompassing shortness of breath, high blood pressure, and bone density. Patient 1 talked for 56 seconds during the 10-minute, 13 second exchange. For the mental health topic, the exchange between Physician A and Patient 1 was as follows: Physician A: “... and you were followed back in behavioral services, and you still see [name of psychiatrist]?” Patient 1: “Yes.” Physician A: “Okay. Things are going well there?” Patient 1: “Yes. It’s going okay.”  When Physician A asked the leading questions, “Things are going well there?” Patient 1 answered “It’s going okay.” “Okay” is a qualified statement and is not the same as “great,” ”excellent,” or “fine.” Conversation analysts have noted that “Okay” in response to an opening exchange typically operates as an invitation for further discussion.

 However, Physician A did not explore why Patient 1’s response was only “Okay.” Therefore, the discussion was considered perfunctory.

The verbal dominance score of Physician B from other visits was 1.24 (ie, she spoke only somewhat more than her patients in the other visits in this sample). Patient 2’s PHQ-2 was 6, the highest score for PHQ-2. Physician B explored empathic opportunities,

 asked 8 of the 9 PHQ-9 questions, diagnosed depression, prescribed Effexor, and made a referral for psychotherapy. Part of the discussion is as follows: Physician B: “And you know what? Sometimes as crappy as it feels that you’re going through all these crappy feelings—” Patient 2: “Oh, this is awful.” Physician B: “&mdash;you could kind of look at it as a gift. Like okay, now’s your time. You have to do it. You have to deal with these things, you know?” Patient 2: “Right. Right.”

Among patients likely needing mental health care, only one-third had evidence-based mental health discussions. Another third of the visits had perfunctory MHDs, leaving the remaining third of patients without any MHD. The significant effects of physician practice styles on presence and quality of an MHD deserve our attention.

Patients seeing physicians who spent more time with their other patients were more likely to have a higher-quality MHD. This finding echoes the observation that “slow medicine” can be more appropriate for serving patients with chronic conditions.

 It is also consistent with previous research that suggests visits in which physicians who provided appropriate counseling or screening took 2.6 to 4.2 minutes longer than visits in which patients did not receive these services.

 An analysis of 190 video-recorded visits in Europe also suggests that when both physician and patient considered psychosocial problems to be important, consultations lasted longer than those about biomedical problems only.

 Nevertheless, some physicians are reluctant to deal with patients’ complex agendas because they are “overly time consuming.”

As our population ages, the number of patients with chronic conditions is rising rapidly. Visit lengths should not be arbitrarily set without much tailoring to patients’ individual needs.

 Offering longer visits for patients with mental health needs would require either smaller caseloads or more staff to do pre-visit and postvisit services, or less frequent visits. Above all, evidence-based practices, such as using standardized mental health assessment tools and engaging patients in shared decision making, ought to be routine and not left to chance.

Efforts to improve quality of care should incorporate evidence beyond the simple association between visit length and quality of care. We need to examine potentially malleable physician communication behaviors and focus on their impact on quality of care. One of those behaviors is verbal dominance. Although it may be more expeditious to actively direct the conversation and maintain control, verbal dominance disempowers patients. Physicians who fear that addressing mental health issues is too time-consuming spend insufficient time addressing their patients’ mental health.

 Other physicians find that asking patients about their suffering and listening to their answers is gratifying and takes little additional time.

 A study of patient clues and physician responses in primary care and surgery found that 76% of patient-initiated clues were emotional in nature, and visits in which physicians missed the opportunity to adequately address patients’ emotional clues were actually longer than visits with a positive response.

 Furthermore, treating mental illnesses can improve the course of comorbid medical illnesses.

 The deficiencies in medical education regarding how to provide evidence-based mental health care should be addressed.

Given that one of the undisclosed agendas for patients scheduling a PHE was to discuss mental health topics, physicians could be more purposeful about eliciting, uncovering, and prioritizing the patient’s agenda. As it is often the case that the most important issues affecting patients’ well-being isn’t always the first topic discussed in a visit, negotiating around time and topics to be discussed becomes critical.

 Agenda setting is difficult; even in a sample of psychiatric encounters, the evidence showed that 2 of every 3 patients were not asked if they had any concerns to discuss.

 Physicians fear that eliciting a complete agenda will be too time-consuming,

 and many physicians also feel unprepared to handle mental health problems.

 Fully eliciting patient concerns adds less than a minute to the visit, however, and teaching these agenda-setting skills requires as little as 3.5 hours to learn and implement.

 Encouraging patients to fully voice their concerns, and preparing physicians to address difficult and potentially uncomfortable topics, can yield more effective consultations

 and mental health discussions, as shown in this study.

The annual costs of PHEs exceeded $10 billion per year, similar to the annual costs of all lung cancer care in the United States.

 Many individuals have called for eliminating annual physicals,

 based on systematic reviews and meta-analyses showing no reduction in morbidity or mortality.

 This study revealed that some patients came to their PHEs with mental health concerns, yet only one-third of patients likely needing mental health services had an evidence-based mental health discussion. Combined with findings of limited application of the 5As (assess, advise, agree, assist, and arrange) in colorectal cancer screening in PHEs,

 this study contributes to the body of evidence that reveals suboptimal quality of care delivered in PHEs. Policy makers should revisit the ongoing challenges of asking primary care physicians, who often are not trained in evidence-based counseling approaches, to deliver this care under increasing time pressures. Doing this poorly is likely a waste of a scarce resource: physician time. We should consider means to offer support outside of the ambulatory encounters so that it is possible to extend the office visit conversation—that often falls short&mdash;beyond the visit so that it approaches recommended counseling content.

Because the Affordable Care Act mandates one “free” PHE per person per year, it is important that PHEs be effective, including identifying patients with mental health needs.  Improving the quality of PHEs may require reimbursement for longer visits, training, and rewards that enable physicians to more fully elicit patients’ agendas and to listen more attentively. Without significant improvement in the quality of PHEs, eliminating them may do more to improving value in healthcare for the nation.

This study included older and privately insured patients in a single integrated delivery organization. The generalizability is not known. An additional limitation is the proxy measure of patient activation (ie, patient bringing a list of issues to the visit). We didn’t have information regarding what topics were on the list. A more specific measure of patient activation related to mental health would have been helpful to understand if mental health was on patients’ agendas. It would have been informative had we been able to document what was on patients’ lists, as it would enable us to understand if there were competing demands for the time with physician. Lastly, the study did not address health literacy training of the physicians and the patients’ ability to be highly verbal. Future research efforts should take account of these important factors.

The cup is one-third full, or two-thirds empty. PHEs could fill an important role for some patients to raise mental health concerns without directly stating that they are having these problems to the scheduling staff. Physicians should be on the look-out to uncover these potential issues and use the time to assess and evaluate mental health (a high-value service), rather than on performing extensive physical exams (low-value activities) during PHEs. Graduate medical education should spend at least as much time on training the next generation of physicians on how to ask open-ended questions, fully elicit patients’ agendas, and listen attentively as on how to listen for heart murmurs. 

 Palo Alto Medical Foundation Research Institute (MT-S, CW, CS, AS), Mountain View, CA; Harvard Medical School (LH, TM), Boston, MA;  Memorial Sloan-Kettering Cancer Center (LD), New York, NY; Indiana University (RF), Indianapolis, IN; Wayne State University (LM), Detroit, MI; University of North Carolina at Chapel Hill (JEL), Chapel Hill, NC.

 National Institute of Mental Health (NIMH R01MH081098) and National Cancer Institute (NCI R01CA112379).

 The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Articles

Among patients likely needing mental health care, two-thirds had no discussion or perfunctory discussion of mental health during periodic health exams.

Periodic Health Examinations and Missed Opportunities Among Patients Likely Needing Mental Health Care

To examine the association between primary carephysician (PCP) reimbursement and delivery of sexually transmitteddisease (STD) services.

Cross-sectional sample of PCPs contracted withMedicaid managed care organizations in 2002 in 8 Californiacounties with the highest rates of Medicaid enrollment andchlamydia cases.

The association between physician reimbursementmethods and physician practices in delivery of STD services wasexamined in multiple logistic regression models, controlling for anumber of potential confounders.

Evidence of an association between reimbursementbased on management of utilization and the PCP practice of providingchlamydia drugs for the partner's treatment was most apparent.In adjusted analyses, physicians reimbursed with capitationand a financial incentive for management of utilization (odds ratio[OR] = 1.63) or salary and a financial incentive for management ofutilization (OR = 2.63) were more likely than those reimbursedunder other methods to prescribe chlamydia drugs for the partner.However, PCPs least often reported they annually screenedfemales aged 15-19 years for chlamydia (OR = 0.63) if reimbursedunder salary and a financial incentive for productivity, or screenedfemales aged 20-25 years (OR = 0.43) if reimbursed under salaryand a financial incentive for financial performance.

Some physician reimbursement methods mayinfluence care delivery, but reimbursement is not consistently associatedwith how physicians deliver STD care. Interventions toencourage physicians to consistently provide guideline-concordantcare despite conflicting financial incentives can maintain quality ofcare. In addition, incentives that may improve guideline-concordantcare should be strengthened.

Referred to as "the captain of the ship," the physiciantraditionally has been the key agent indelivery of healthcare services.

Many reforms infinancial and organizational incentives have beenimplemented to control healthcare expenditures andimprove quality of care by influencing physicians' practice.Analysts have been tracking how financial andorganizational incentives in the healthcare systemaffect physician practice for decades and they have documentedthat physicians' practice is related to theincentives in the payment system.

Recent evidenceshows a decreasing trend, from 1997 to 2001, in thepercentage of physicians who were subject to reimbursementbased on profiling, patient satisfaction, orquality measures. In 2001, 17% of physicians reportedthat rates of preventive-services screening were used indetermining their compensation. In contrast, the percentageof physicians who are encouraged to followtreatment guidelines has increased.

The impact of different types of reimbursement onphysician practices has been examined in some detail.The Institute of Medicine reported that payment incentivesare misaligned with the delivery of high-qualitycare.

The methods of physician payment (eg, fee-for-service[FFS], salary, capitation) usually do not rewardgood patient health outcomes. For instance, FFS offersincentives for providing more services, whereas capitationrewards conserving resources. One can result inoveruse and the other in restricted use of resources.Although salary-based compensation does not necessarilyencourage underuse or overuse, it does not provideincentives for improved productivity or efficiency. Noneof the payment systems, unless used in conjunctionwith other strategies, provide incentives for high quality.In some cases, perverse incentives could rise againstquality improvement.

Evidence shows that capitated physician groups putmore emphasis on monitoring overuse than underuse ofservices like immunization, which may be attributed tofinancial incentives of capitation.

Others have foundthat physician incentives that are based on the physician'sown production (rather than the group) increasephysician productivity.

Some financial-productivityincentives may discourage the delivery of preventivecare such as Pap smears and cholesterolchecks, but not mammogramsand flu shots.

Clinical practice guidelineshave been developed for manychronic diseases to promote practicesthat are consistent with currentscientific understanding ofthe disease and treatment modalities,and to provide scientific guidancethat could lead to optimalpatient health outcomes. For sexuallytransmitted diseases (STDs),guidelines are primarily developedby the Centers for DiseaseControl and Prevention

In California, state lawsand the California ChlamydiaAction Coalition

have providedadditional practice guidelines. Incombination, these guidelines aredesigned to promote comprehensiveand effective STD care.Evidence suggests that productivity,quality, and cost-containmentincentives are associatedwith use of practice guidelines.

Self-reported data indicate thatfor more than 50% of physiciansstudied, treatment guidelines havea moderate to very large effect ontheir practice.

Although practice guidelinespresent a potential solution to thecomplexity of medical decisions,

evidence of adherence to guidelinesfor various diseases has beendiscouraging.

In addition,research on the impact of financialincentives on guideline adherenceis limited. A small-scalerandomized, controlled trial suggestedthat FFS physicians providedmore care and conformed topediatric care guidelines, whereas salaried physicianswere in less conformity with guidelines.

However, thecurrent literature provides little information about theimpact of reimbursement (and the incentives reimbursementrepresents) on delivery of STD care.

In this study, we investigate the association betweenphysician compensation—independent of other determinantsof physician practice—and the delivery ofguideline-concordant STD care. The STD guidelines weexamined included taking a sexual history of the patientat the first nonurgent visit,

annual screening of sexuallyactive females 15 to 25 years of age,

providingchlamydia drugs for the partner's treatment orpatient-delivered partner therapy,

and providing servicesto minors without parental notification or consent(also California State Law, Family Code §6926(a)).

Surveys of Primary Care Physicians (PCPs) contractedwith Medicaid HMOs in 2002 in 8 California countieswith the highest rates of chlamydia and Medicaid HMOenrollment were used for this study. The PCPs who participatedin the survey were contracted with a total of 25Medicaid HMOs in the selected counties. Three plansfolded early in the study period, and their PCPs either recontractedwith another HMO in the study or lost theirMedicaid HMO contracts. Two plans with fewer than10 000 enrollees each refused to participate in the study.An electronic version of the participating plans' PCPdirectory was obtained, and an unduplicated database ofall PCPs contracted with these HMOs was constructedand served as the sampling frame.

These PCPs were contacted by phone up to 12 timesfor a telephone interview from January through May2002 and were offered $75 to participate in the 15-minute survey. The choice of a self-administered surveywas offered to those PCPs who were unable to completethe interview by phone. The adjusted response rate was41% (948) following the methodology used in anothernational survey of physicians.

Of the unduplicated listof physicians provided by participating HMOs, 64%(6096) were found to be ineligible primarily due to outdatedcontact information or contractual changes in thetime between the collection of the PCP list from HMOsand the fielding of the survey, and secondarily due tobeing specialists outside the scope of the study.Additional analysis of respondents and nonrespondentson the basis of available characteristics, including countyand specialty, did not identify any nonresponse biasin the sample. This study was approved by the appropriateinstitutional review board, and all study subjectsconsented to participate in the survey.

Primary care physician adherence to STD guidelineswas captured on a 5-point Likert scale ranging from 1 to5, with 1 representing "always," 2 representing "usually,"3 representing "sometimes," 4 representing "rarely,"and 5 representing "never." For the following analyses,all adherence variables were dichotomized into thosewho consistently (always, usually) followed a guidelineversus those who did not (sometimes, rarely, never).This decision was based on the assumption that adherenceto any practice guidelines often depends on thephysician judgment of the appropriateness of the treatmentor procedure given the presentation of illness,the patient's characteristics, and other circumstances.Thus, consistent adherence to guidelines can be appropriatelydefined as "always" and "usually" followingguidelines.

The main independent variable was PCP paymentmechanism. Primary care physicians were asked toidentify whether they were salaried physicians of anHMO or a medical group. Those who were not salariedwere then asked whether they were reimbursed on acapitation or FFS basis by their affiliated health plan ormedical group. They also were asked whether they contracteddirectly with the health plan, or through themedical group that provided the largest proportion oftheir Medicaid HMO patients. Primary care physiciansalso were asked whether their contracts with their HMOsor medical groups included stipulations for reimbursementbased on their productivity (eg, number of visits),quality of care (eg, patient satisfaction or peer review),their management of utilization (eg, rate of referrals, laboratorytests, x-rays), or financial performance of thegroups (eg, profit sharing). Each PCP may have reportedmore than 1 type of reimbursement.

Nine dichotomous independent variables were createdto distinguish those PCPs reimbursed on an FFS, capitation,or salary basis and each of the 4 contractualstipulations or reimbursement mechanisms. The percentageof those reimbursed under FFS and any of the 4reimbursement mechanisms was too small for thesegroups to be separately identified (ranging between 5.6%to 3%). The remaining variables included the 2 paymentmethods (capitation and salary) crossed with the 4 reimbursementmechanisms (productivity, quality of care,management of utilization, and financial performance(

). Many PCPs reported more than 1 reimbursementmechanism, so the resulting variables representmutually exclusive payment methods but overlappingreimbursement mechanisms. Therefore, the analyses ofinfluence of payment on PCP practice are interpreted,for example, as the influence of salary and productivityon PCP practice versus the influence of other paymentmethods.

A number of other factors that could affect PCPadherence to STD practice guidelines were controlled forin the logistic regression models. These included businesscharacteristics of PCPs (practice setting, volume ofMedicaid patients in practice, number of Medicaid HMOcontracts, and number of medical group contracts withMedicaid business), personal characteristics (sex, specialty,and years in practice), having STD guidelinesfrom the Centers for Disease Control and Prevention andUS Preventive Services Task Force, having ever receivedfeedback on STD screening from the contractedMedicaid HMO or medical group, and the type of contractedMedicaid managed care health plan (2-planmodel: 1 commercial and 1 local initiative plan for beneficiariesto choose from; county organized health system:an agency organized and operated by the county;and geographic managed care: capitated contractsbetween the state and multiple commercial plans for ageographic area).

We examined the extent to which PCPs followedexisting STD guidelines given PCP reimbursement. Theassociation between PCPs' adherence to STD practiceguidelines and reimbursement (as well as betweenadherence and control factors) was assessed using chisquaretests of difference and reported at 

< .05. Therelationship between payment and adherence to STDpractice guidelines was assessed in a series of logisticregression models controlling for confounding factors.Each model was adjusted for clustering of PCPs withinHMOs, using STATA v.7.0.

The analyses were notweighted otherwise because the sample frame includedthe universe of PCPs who contracted with the MedicaidHMOs in the selected counties.

Because of the cross-sectional nature of the data, wecannot accurately determine whether there is a causalrelationship between the payment-mechanism variablesand the outcomes. It is possible that physicians simultaneouslychoose their practice based on the paymentmechanisms used in that practice. Such physicians alsomay have characteristics that make them more or lesslikely to provide recommended STD care. For example,a physician who has a predisposition to follow STDguidelines may prefer to join a group that pays itsproviders on an FFS basis. Our data do not allow us todetermine whether it is the reimbursement variables,or alternatively, unmeasured characteristics of thephysician, that cause the physician to provide thegiven STD care.

To investigate this issue further, we conducted amultinomial logistic regression where payment mechanism(capitation, FFS, or salary) was the dependentvariable and various physician and practice characteristicswere the independent variables. Several of thesevariables were found to be statistically significant; togive a single example, physicians who graduated fewerthan 10 years ago were more likely to be salaried, and itis plausible that more recently trained physicians aremore likely to follow guidelines due to their familiaritywith current STD guidelines. Similarly, one mightexpect several unmeasured characteristics to have similarcorrelations. Ideally, one would use instrumentalvariablestechniques to form exogenous predictors ofpayment mechanisms, but the dataset used does notcontain the necessary variables that can predict paymentmechanism but are unrelated to the provision ofSTD care. As a result, the analysis below does not drawconclusions asserting that the payment mechanismshave a causal relationship with provision of care.Rather, only a relationship between the 2 sets of variablesis posited.

The most frequent method of reimbursementreported was salary and quality of care (22%), followedby FFS (20%), salary and productivity (18%), salary andfinancial performance (18%), and capitation and managementof utilization (16%) (Table 1). The least commonlyused method was capitation and productivity(8%). A large proportion of PCPs (43%) reported seeing10 or fewer female (15-25 years old) Medicaid patientsper week, and a similarly high proportion (44%) reportedthat their practice consisted of 11% to 50% Medicaidpatients. About 2 in 5 (41%) were in solo practice, andless than half (48%) contracted with only 1 HMO or 1medical group (42%). Thirty-five percent of PCPs werefemale; 40% were family or general practitioners, 25%were internists, and 28% were pediatricians.

Forty-three percent had STD guidelines from theCenters for Disease Control and Prevention or the USPreventive Services Task Force. Twenty-one percenthad ever received feedback on their STD screeningpractices from their affiliated HMOs or medical groups.Most (69%) were affiliated with HMOs in counties withthe 2-plan Medicaid managed care model.

Adherence to guidelines varied by the type of guideline.Primary care physicians most often reported consistentlyobtaining sexual history at the first nonurgentvisit (68%) and screening young females 15-19 (59%)and 20-25 (62%) years of age (

). Primary carephysicians least often reported consistently providingchlamydia drugs for the partner's treatment (36%) or providingservices to minors without parental or guardiannotification or consent (48%).

The type of reimbursement was significantly associatedwith adherence to 2 guidelines. Those reimbursedunder salary and quality of care or salary and managementof utilization more often adhered to annualscreening of sexually active females aged 15-19 years(67%, 

< .05) than those reimbursed under other methods.Those reimbursed under capitation and quality ofcare more often consistently screened females aged 20-25 years for chlamydia annually (52%, 

< .05) thanthose reimbursed under other methods. Provision ofchlamydia drugs for the partner's treatment was moreconsistently done by PCPs who were reimbursed undercapitation and management of utilization (41%, 

< .05)or salary and management of utilization (45%, 

< .001)than those reimbursed under other methods.

Some reimbursement mechanisms were related tophysicians' adherence to 2 guidelines, after accountingfor other confounders (

). Salary and productivitywere associated with a higher likelihood of consistentchlamydia screening of sexually active females aged 15-19 years annually (odds ratio [OR] = 0.63, 

< .05),whereas salary and financial performance were associatedwith a lower likelihood of consistent chlamydiascreening of females aged 20-25 years annually (OR =0.43, 

< .001). Alternatively, the likelihood of providingchlamydia drugs for the partner's treatment was higher ifthe PCP was reimbursed through capitation and managementof utilization (OR = 1.63, 

< .05) or salary andmanagement of utilization (OR = 2.63, 

< .05). Overall,4 out of 40 reimbursement mechanisms examinedshowed statistically significant associations with guideline-concordant practices. Only 2 of the 40 reimbursement mechanisms would have been significant at the 5%level by chance.

In this study, no clear or consistent patterns emergedindicating a link between methods of PCP payment andPCP delivery of all STD care. Our results do suggest thatthere may be some associations between reimbursementand delivery of certain guideline-concordant STD services.Financial incentives imbedded in management of utilizationmay influence the PCP's decision to provide thepatient with chlamydia medication for the partner'streatment. This guideline is intended to curb the spreadof chlamydia infection and its costly complications inwomen through treating the partner, overcoming thenumerous barriers facing the partner. Treating the partnercan ensure that the patient is cured and will notreturn for additional visits or care because of reinfectionsand further complications. Curtailing repeated visitscan reduce rates of use of physician services, which isthe primary goal of management of utilization.

The financial incentives of salaried PCPs with productivityor financial-performance stipulations in theircontracts provide disincentives for the annual chlamydiascreening recommended by the guideline.Chlamydia testing requires longer appointments to discusssexual-risk history and possibly to collect a specimen,thereby incurring additional lab costs. Salary andproductivity goals could discourage such practices.

The lack of a clear association between reimbursementmechanisms and the outcome variables may bedue to the fact that physicians may contract with anumber of commercial and Medicaid HMOs and medicalgroups who may or may not encourage physicians toadhere to certain guidelines. In this study, we examinedthe relationship between guideline adherence and paymentby the HMO and medical groups that provided thelargest proportion of Medicaid patients to the physician'spractice. The collective impact of various reimbursementmechanisms may paint a different picture ofPCP adherence to STD guidelines.

Evidence is emerging about the positive impact ofperformance contracting on various patient outcomes,

although we did not examine the effect ofsuch techniques on physicians' adherence to guidelines.However, factors such as multiple contracts maydilute the impact of incentives, and physicians' lack ofrecall of the existence of incentives from certain planscan reduce the effectiveness of such incentives.

Furthermore, monitoring physician performance canbe a costly proposition.

Additional research is clearlynecessary to better understand the relationshipbetween reimbursement mechanisms and physicians'provision of guideline-concordant care or best practices.If reimbursement mechanisms indeed contradictguidelines, modifications of such mechanisms in PCPcontractual agreements are needed. Alternatively,reimbursement mechanisms that may jointly increasethe positive impact of best-practice guidelines can beincorporated in organizational quality improvementefforts to maximize their utility.

This study included only PCPs contracted withCalifornia's Medicaid HMOs and excluded PCPs whoprovided care to Medicaid patients outside the managedcare setting. Thus, the findings of this study areapplicable to Medicaid HMOs only and cannot be comparedwith FFS Medicaid. Medicaid HMO providers maydiffer from FFS Medicaid providers in their delivery ofSTD care, and California's Medicaid managed careproviders may differ from those in other states.Furthermore, all data were self-reported, and physicianshave been found to self-report guideline-concordantpractices at a consistent, but higher, rate than therate indicated by chart reviews.

In addition, smallsample sizes limited our ability to draw further conclusionsregarding the potential impact of reimbursementon PCP adherence. Finally, as discussed earlier, causalitycannot be inferred from the multivariate analysisbecause physicians may self-select into practices thathave particular payment arrangements.

Author | Ming Tai-Seale, PhD

Articles

Periodic Health Examinations and Missed Opportunities Among Patients Likely Needing Mental Health Care

Association Between Physician Compensation Methods and Delivery of Guideline-Concordant STD Care: Is There a Link?