Solicited reports of adverse events included redness, swelling, and pain at the injection site, as well as fatigue, fever, headache, and malaise. The HSV vaccine was more reactogenic and was associated with local pain, redness, and swelling more often than was the control vaccine. There was a small but significant increase in systemic symptoms, including fatigue, fever, headache, and malaise, in the HSV-vaccine group (Table 3). Dose 2 and dose 3 of the HSV vaccine were not associated with increased reports of adverse events; in contrast, reactogenicity decreased slightly with additional vaccination.

    Immunogenicity

    The HSV vaccine was immunogenic and stimulated ELISA and neutralizing antibodies. As expected, control subjects did not have antibody to gD-2 on ELISA or neutralization of HSV-2. Geometric mean gD-2 ELISA titers were 21 at baseline and 6809 at month 7 after three doses of HSV vaccine; ELISA titers waned to 769 by month 20. HSV-2 neutralizing antibodies developed after two doses of HSV vaccine, but the median value fell to an undetectable level by study month 6 (see the Supplementary Appendix for details). After dose 3, HSV-2 neutralizing antibodies were again above the limit of detection (the mean titer at month 7 was 29) but fell to a median value that was undetectable by study month 16.

    Discussion

    Our findings of vaccine efficacy against HSV-1 and lack of efficacy against HSV-2 are puzzling in view of the previous two studies involving discordant couples that showed efficacy of this gD-2 vaccine against HSV-2.4 The difference in efficacy is likely to be due to some factor in the two populations studied. The distinguishing feature of discordant couples is that they are a highly selected group in which the uninfected partner is potentially repeatedly exposed to HSV by the infected partner. Attack rates of HSV-2 genital disease in the prior studies of gD-2 vaccine were high among uninfected women in discordant couples (13.9% for 19 months or 8.4% per year) and were reduced significantly by the vaccine (efficacy, 73% and 74% in the two trials; P=0.01 and P=0.02, respectively).4 Too few cases of HSV-1 genital disease occurred in women in the two previous studies to assess efficacy against HSV-1 (one case in each study among women who were seronegative at study entry).4 Potential reasons for vaccine efficacy in the discordant-couple population include selection bias for a population of women with relative resistance to HSV-2, with added benefits from a subunit vaccine; an undefined immunologic priming event from chronic sexual exposure to HSV-2 viral antigens from the infected partner; and less frequent sexual activity due to the long-term nature of the relationship as compared with sexual activity by couples in new relationships.

    It is not apparent why the biologic characteristics of HSV-1 are different from those of HSV-2; the gD-2 vaccine induces significant protection against genital HSV-1 disease as well as HSV-1 infection, but not against disease or infection caused by HSV-2. Genital HSV-1 may be acquired primarily through oral–genital sex (although a history of oral sex was not a risk factor for HSV-1 acquisition in our study); a lower inoculum, an oral–genital route (possibly less traumatic sex), and a less suitable environment for HSV-1 replication are all possible explanations for protection by the vaccine against HSV-1 but not HSV-2. The gD-2 antigen is derived from HSV-2, but 89% amino acid homology is shared with gD-2 from HSV-1, which may explain the protection against HSV-1. Type-specific immune responses to vaccine antigen may reveal differences in antibody activity against HSV-1 and HSV-2. Whether HSV-1 is more easily neutralized by vaccine-induced antibodies must be determined by further laboratory studies.

    Self-reported behaviors and demographic factors were analyzed for association with HSV-1 or HSV-2 infection in the per-protocol cohort for months 2 through 20. Analysis of self-reported behavioral risk factors was restricted to 5980 sexually active subjects. An increased risk of HSV-1 infection was associated with 6 or more lifetime sexual partners (hazard ratio, 2.2; 95% CI, 1.3 to 3.8) and more than 1 partner in the previous 12 months (hazard ratio, 1.9; 95% CI, 1.4 to 2.7). Subjects who were 23 years of age or older were less likely to acquire HSV-1 than 18-to-22-year-olds (hazard ratio for subjects 23 to 26 years of age, 0.6; 95% CI, 0.4 to 0.8; hazard ratio for subjects 27 to 30 years of age, 0.4; 95% CI, 0.3 to 0.8). Factors not associated with an increased risk of HSV-1 infection included race or ethnic group, country of residence (United States or Canada), having a current partner with herpes, ever having a partner with herpes, condom use, history of any sexually transmitted infection (STI), and oral sex.

    An increased risk of HSV-2 infection was associated with having 6 or more lifetime sexual partners (hazard ratio, 2.0; 95% CI, 1.1 to 3.8), having 6 or more partners in the previous 12 months (hazard ratio, 2.7; 95% CI, 1.3 to 5.5), ever having a partner with herpes (hazard ratio, 3.0; 95% CI, 1.7 to 5.3), having a current partner with herpes (hazard ratio, 3.4; 95% CI, 1.8 to 6.4), a history of any STI (hazard ratio, 3.3; 95% CI, 2.2 to 5.0), nonwhite race (hazard ratio, 3.1; 95% CI, 2.1 to 4.6), and U.S. residence (hazard ratio, 2.7; 95% CI, 1.2 to 6.2). Factors not associated with increased risk of HSV-2 infection included age, ethnic group, condom use, and oral sex. Initiation of sexual activity after 15 years of age was associated with a decreased risk of both HSV-1 infection (hazard ratio for initiation at 16 to 18 years of age, 0.6; 95% CI, 0.4 to 0.8; hazard ratio after 18 years of age, 0.3; 95% CI, 0.2 to 0.6) and HSV-2 infection (hazard ratio for initiation at 16 to 18 years of age, 0.5; 95% CI, 0.3 to 0.8; hazard ratio after 18 years of age, 0.3; 95% CI, 0.2 to 0.6).

    Genital Shedding of HSV-2

    Forty-three subjects (30 in the HSV-vaccine group and 13 in the control group) with HSV-2 infection collected anogenital swabs on 60 consecutive days, beginning 3 to 6 months after disease onset (15 subjects in the HSV-vaccine group and 9 in the control group) or seroconversion (15 subjects in the HSV-vaccine group and 4 in the control group). Analysis of these swabs showed that the rate of viral shedding was higher among the HSV-vaccine recipients than among controls (29% vs. 19%; relative risk, 1.55; 95% CI, 1.28 to 1.86). The mean quantity of HSV DNA on days with shedding did not differ between the two groups.

    Statistical Analysis

    The trial was designed to have 80% power to detect a vaccine efficacy of 75% with 45% as the lower limit of the 95% confidence interval. It met the information goal, observing 70 of a planned 72 cases of genital herpes disease in the per-protocol cohort. The trial was monitored by an independent data safety and monitoring board sponsored by the National Institute of Allergy and Infectious Diseases, which met quarterly and reviewed the study for safety. At a prespecified interim analysis, the board also reviewed the trial for futility. The sample size was extended once in response to higher-than-anticipated attrition. Vaccine efficacy was estimated as 1 minus the relative risk from a Cox proportional-hazards model fit to the time to first acquisition of each study end point. Rates of loss-to-follow-up were similar between the two study groups, and noninformative censoring was assumed. A post hoc assessment of demographic and behavioral risk factors for HSV acquisition was performed with the use of a Cox proportional-hazards model adjusted for the receipt of HSV vaccine. All reported P values are two-tailed and have not been adjusted for multiple testing. The per-protocol and intention-to-treat cohorts are defined in the legend for Figure 1.
    Results
    Characteristics of the Study Population

    Fifty clinical sites in the United States and Canada screened a total of 31,770 women for antibodies to HSV-1 and HSV-2; 12,468 women were seronegative for both HSV-1 and HSV-2, of whom 8323 met the other eligibility criteria and were enrolled between January 14, 2003, and November 19, 2007.

    Vaccine Efficacy

    In the control group, HSV-1 was a more common cause of genital disease than HSV-2 (21 cases caused by HSV-1 vs. 14 cases caused by HSV-2). Efficacy against genital disease caused by HSV-1 was observed (vaccine efficacy, 58%; 95% CI, 12 to 80) (Figure 2B), but efficacy was not observed against HSV-2 disease (−38%; 95% CI, −167 to 29) (Figure 2C). Three doses of vaccine were associated with efficacy against HSV-1 (77%; 95% CI, 31 to 92) but not HSV-2 (−40%; 95% CI, −234 to 41). An analysis in which the case definition was limited to culture-positive cases (excluding HSV cases diagnosed according to clinical and serologic criteria) also showed efficacy against HSV-1 (two-dose efficacy, 69%; 95% CI, 25 to 87; three-dose efficacy, 82%; 95% CI, 35 to 95).

    The HSV vaccine provided protection against infection caused by HSV-1 or HSV-2 (efficacy, 22%; 95% CI, 2 to 38). This overall finding of protection against infection was driven by efficacy against HSV-1 infection (35%; 95% CI, 13 to 52), whereas efficacy against HSV-2 infection was not observed (−8%; 95% CI, −59 to 26).

    The control vaccine, inactivated hepatitis A vaccine (Havrix, GlaxoSmithKline), was formulated as 720 enzyme-linked immunosorbent assay (ELISA) units of inactivated hepatitis A virus combined with 0.5 mg of alum, in a volume of 0.5 ml. For the study to be blinded, the control vaccine was given at 0, 1, and 6 months in doses containing one half the usual volume and one half the usual amount of antigen.

    Study Design

    The double-blind, randomized field trial was designed in collaboration among the trial sponsors, the National Institutes of Health (NIH) and GlaxoSmithKline; the study chair; the executive committee; and the scientific leadership group. Data were collected with the use of the GlaxoSmithKline remote data-entry system and were monitored by GlaxoSmithKline. All the authors and the trial sponsors vouch for the accuracy and completeness of the data. Data were electronically transferred to EMMES, a contract research organization, where they were analyzed according to the analysis plan prepared by biostatisticians at GlaxoSmithKline and EMMES, with input to the analysis plan from the NIH, the study chair, and the executive committee. The manuscript was drafted by the first author, with input from the biostatisticians at EMMES and from the executive committee.

    Study End Points

    The primary end point of the study was prevention of genital herpes disease caused by HSV-1, HSV-2, or both from month 2 (1 month after vaccine dose 2) through month 20. Genital disease was defined as clinically compatible signs and symptoms confirmed by viral culture, seroconversion, or both within 6 months after disease onset. Secondary end points included prevention of HSV-1 or HSV-2 infection (with or without disease) from month 2 through month 20 (two-dose efficacy) or month 7 through month 20 (three-dose efficacy) and prevention of genital herpes disease caused by individual HSV types. Cases of infection and disease were determined centrally by an independent, blinded end-point review committee with the use of documented criteria.

    Substudy of Viral Shedding

    Subjects identified as having acquired genital HSV-2 disease or HSV-2 infection during the study were invited to participate in an evaluation of viral shedding. Subjects were instructed to collect daily swabs from the anogenital area for 60 consecutive days and to maintain a diary of genital signs and symptoms, as previously described, beginning 3 to 6 months after HSV-2 seroconversion or disease onset.

    Laboratory Studies

    Western blot analysis (University of Washington Clinical Virology Laboratory at Seattle Children’s Hospital) was used to confirm HSV-1– or HSV-2–seronegative status at study entry and seroconversion during the follow-up period.6 Seroconversion to HSV-1 or HSV-2 was defined as a positive Western blot analysis in a subject with a previously negative analysis for the corresponding HSV type.

    Serum specimens from a random subset of 611 subjects in the HSV-vaccine group and 223 subjects in the control group were assessed for the development of antibodies to vaccine antigens with the use of ELISA4,7 for gD-2 and virus neutralization of HSV-2 at 0, 2, 6, 7, 12, 16, and 20 months. Long-term genital shedding of HSV-2 DNA was assessed with the use of a quantitative, real-time, fluorescence-based polymerase-chain-reaction assay, as described previously, with a positive result defined as 150 copies per milliliter.

    Both herpes simplex virus type 1 (HSV-1) and type 2 (HSV-2) can cause primary infection of the genital tract, and HSV-1 infection has become an increasingly frequent cause of genital disease. The majority of HSV infections are asymptomatic, and only 10 to 25% of persons with HSV-2 antibodies have recurrent genital disease. Transmission of HSV from infected women to neonates may lead to severe neurologic disease or death in the newborn. Strategies to control genital herpes infection and disease have mainly focused on antiviral chemotherapy, education, and the use of condoms. The availability of an effective prophylactic vaccine would help control genital herpes.

    In two previous efficacy trials of an HSV-2 glycoprotein D–based subunit (gD-2) vaccine in discordant couples in which one partner had recurrent HSV genital disease, the subset of seronegative women (negative for both HSV-1 and HSV-2 antibodies) was significantly protected against HSV-2 disease by the vaccine (73% and 74% efficacy, respectively); efficacy was not shown in either men or HSV-1–seropositive women.4 To further evaluate the gD-2 vaccine as a potential public health tool, we evaluated this vaccine in a cohort of women who were screened and found to be antibody-negative for HSV-1 and HSV-2.

    Study Population

    Women 18 to 30 years of age who were seronegative for HSV-1 and HSV-2 were recruited from 40 sites in the United States and 10 in Canada. Other inclusion criteria were written informed consent, the absence of serious health problems, a willingness to use effective methods of birth control throughout the period from 30 days before vaccination to 2 months after receipt of the third dose of vaccine, and a negative pregnancy test.

    Vaccines and Adjuvants

    The HSV-2 vaccine (GlaxoSmithKline) contained 20 μg of truncated glycoprotein D from HSV-2 strain G. The gD-2 vaccine antigen was combined with an adjuvant that consisted of 0.5 mg of aluminum hydroxide (alum) and 50 μg of 3-O-deacylated monophosphoryl lipid A. The vaccine was administered by injection at a dose of 0.5 ml at 0, 1, and 6 months.

    The fourth and final habit is self-study. Beyond ensuring that their clinical practices are consistent with the most recent science, these organizations also examine positive and negative deviance in their own care and outcomes, seeking new insights and better outcomes for their patients.5 By contrast, most health care organizations treat clinical knowledge as a property of the individual clinician, “managing” knowledge only by hiring and credentialing competent professional staff.

    High-value organizations treat clinical knowledge as an organizational as well as individual property. They create knowledge and innovations with the use of some common tools (sentinel-event reporting and root-cause analysis) and some less common ones (monitoring of protocol overrides and rapid-cycle experimentation). Some have units — for instance, the Mayo Clinic’s See-Plan-Act-Refine-Communicate (SPARC) program — that are dedicated to developing innovations in-house, and most have academies to teach leaders and staff the principles and techniques for improving the value of care and to support the application of these principles to high-priority clinical programs and processes. Most important, these organizations deliberately nurture a culture that supports learning by encouraging dissenting views and overriding of specified clinical decision rules (habit 1).

    These habits are not unique to high-value health care organizations. Many delivery organizations engage in some of them — designing clinical pathways and reporting on quality and safety, for instance. But high-value organizations are distinct in two important ways. First, they engage in all four habits systematically. For them, these activities are truly habits, baked into their structures, culture, and routines, not simply short-lived projects. Second, the habits are integrated into a comprehensive system for clinical management that is focused more on clinical processes and outcomes than on resources. A consensus is emerging about how to manage clinical care.

    Each organization expresses these four habits differently. Each faces a unique regulatory and reimbursement environment and has different resources, so each uses different tools and terminologies, varying in the details of how they specify decisions or measure clinical processes. Still, the habits are the same. As we seek models for achieving high-value health care, we must look past the particularities of local structures and tactics to the habits they reflect. Although a “dominant” delivery model may not be transferrable, the habits of high-value health care may be.

    The specification of choices, transitions, subgroups, and patient pathways represents a substantial investment in advance planning. It contrasts sharply with the common practice of focusing management planning on the utilization of expensive resources, such as tests, procedures, and bed-days, rather than on the problems these resources are designed to solve. Many hospitals and clinicians do not plan care processes in advance in such detail; instead, they treat each new patient or problem as a random draw from a heterogeneous population and must therefore reinvent the strategy for solving it.

    A second common habit is infrastructure design. High-value health care organizations deliberately design microsystems3 — including staff, information and clinical technology, physical space, business processes, and policies and procedures that support patient care — to match their defined subpopulations and pathways. Thus, different conditions or patient groups have different microsystem designs. The various tasks of care are allocated to different members of a clinical team (including the patient), with the skill and training of each staff member matched to the work. Such organizations make thoughtful use of assistive personnel and alternative providers, and they ensure that each has the necessary resources by carefully designing the supply chain of equipment and information, simplifying workflow, and reducing work stress. They also harmonize the parts of their management system so that budgets, incentives, data, goals, clinical processes, educational programs, and team structures are all mutually reinforcing.4 Unit-level routines, such as joint ward rounds, team meetings, and executive “walk-arounds,” help tie microsystem components together.

    Attention to microsystem design and integration represents an important shift away from general-services-organization designs that use a single platform to meet the needs of many different patient groups and that focus on maximizing the use of scarce resources, such as operating-room slots, ICU beds, and physicians.

    The third habit is measurement and oversight. For many, measurement of clinical operations is driven by external audiences: payers, regulators, and rating agencies. Although high-value organizations share this reporting obligation, they primarily use measurement for internal process control and performance management. They collect more (and more detailed) measurements than those required for external reporting, selecting those that inform staff about clinical performance. For instance, of the 200-plus measurements used by Intermountain, more than two thirds were developed or refined internally rather than imported unmodified from external agencies. Moreover, such organizations integrate their measurement activities with other organizational priorities such as pay for performance, annual target setting, and improvement activities, making measurement an integral part of accountability and performance management. For example, each year Intermountain’s board selects a different group of measurements from the institution’s overall measurement set to use for annual quality and efficiency bonuses.

    Recent attention to the question of value in health care — the ratio of outcomes to long-term costs — has focused on problems of definition and measurement: what outcomes and which costs? Less attention has been given to an equally difficult but important issue: how do health care delivery organizations reliably deliver higher value?

    It would certainly simplify health care reform if we could show the superiority of a dominant delivery model (e.g., the accountable care organization or the medical home) and roll it out nationwide, developing and proving new approaches to creating value only once. However, experience suggests that not only do new delivery models — for example, integrated networks — not necessarily live up to their promise, but they are surprisingly difficult to transfer, even when successful; those that succeed in one U.S. region haven’t always done well in another. Organizations considered to be among the nation’s highest performers, such as the members of the new High Value Healthcare Collaborative, often have unique personalities, structures, resources, and local environments. Given the health care sector’s mixed record of disseminating clinical innovations and system improvements, how do we learn from leading organizations?

    Although high-value health care organizations vary in structure, resources, and culture, they often have remarkably similar approaches to care management. Specific tactics vary, but their “habits” — repeated behaviors and activities and the ways of thinking that they reflect and engender — are shared. This is important because experience suggests that such habits may be portable.

    The first common habit is specification and planning. To an unusual extent, these organizations specify decisions and activities in advance. Whenever possible, both operational decisions, such as those related to patient flow (admission, discharge, and transfer criteria), and core clinical decisions, such as diagnosis, tests, or treatment selection, are based on explicit criteria. Criteria-based decision making may be manifest in the use of clinical decision support systems and treatment algorithms, severity and risk scores, criteria for initiating a call to a rapid-response team or triggering the commitment of a future resource (e.g., a discharge planner, preprocedure checklists, and standardized patient assessments), and for patients, shared decision making.

    Specification also applies to separating heterogeneous patient populations into clinically meaningful subgroups — by disease subtype, severity, or risk of complications — each with its own distinct pathway. For example, Dartmouth’s Spine Center uses a detailed intake assessment that combines the 36-Item Short-Form Health Survey, computerized visual aids, and shared decision making to sort patients according to the likelihood that they will do better with either medical or surgical care. Similarly, genomic testing has allowed oncology units to divide patients into separate groups according to their probable response to specific therapies (for instance, KRAS testing for cetuximab therapy). And at Intermountain Healthcare in Utah and Idaho, the needs of psychiatric patients are divided into mild (routine care by a primary care physician), moderate (team care), and severe (specialist referral), with a scoring system based on published guidelines. Some organizations, such as Children’s Hospital Boston, are developing standard approaches to uncommon and complex conditions.

    The fourth and final habit is self-study. Beyond ensuring that their clinical practices are consistent with the most recent science, these organizations also examine positive and negative deviance in their own care and outcomes, seeking new insights and better outcomes for their patients.5 By contrast, most health care organizations treat clinical knowledge as a property of the individual clinician, “managing” knowledge only by hiring and credentialing competent professional staff.

    High-value organizations treat clinical knowledge as an organizational as well as individual property. They create knowledge and innovations with the use of some common tools (sentinel-event reporting and root-cause analysis) and some less common ones (monitoring of protocol overrides and rapid-cycle experimentation). Some have units — for instance, the Mayo Clinic’s See-Plan-Act-Refine-Communicate (SPARC) program — that are dedicated to developing innovations in-house, and most have academies to teach leaders and staff the principles and techniques for improving the value of care and to support the application of these principles to high-priority clinical programs and processes. Most important, these organizations deliberately nurture a culture that supports learning by encouraging dissenting views and overriding of specified clinical decision rules (habit 1).

    These habits are not unique to high-value health care organizations. Many delivery organizations engage in some of them — designing clinical pathways and reporting on quality and safety, for instance. But high-value organizations are distinct in two important ways. First, they engage in all four habits systematically. For them, these activities are truly habits, baked into their structures, culture, and routines, not simply short-lived projects. Second, the habits are integrated into a comprehensive system for clinical management that is focused more on clinical processes and outcomes than on resources. A consensus is emerging about how to manage clinical care.

    Each organization expresses these four habits differently. Each faces a unique regulatory and reimbursement environment and has different resources, so each uses different tools and terminologies, varying in the details of how they specify decisions or measure clinical processes. Still, the habits are the same. As we seek models for achieving high-value health care, we must look past the particularities of local structures and tactics to the habits they reflect. Although a “dominant” delivery model may not be transferrable, the habits of high-value health care may be.

    The specification of choices, transitions, subgroups, and patient pathways represents a substantial investment in advance planning. It contrasts sharply with the common practice of focusing management planning on the utilization of expensive resources, such as tests, procedures, and bed-days, rather than on the problems these resources are designed to solve. Many hospitals and clinicians do not plan care processes in advance in such detail; instead, they treat each new patient or problem as a random draw from a heterogeneous population and must therefore reinvent the strategy for solving it.

    A second common habit is infrastructure design. High-value health care organizations deliberately design microsystems3 — including staff, information and clinical technology, physical space, business processes, and policies and procedures that support patient care — to match their defined subpopulations and pathways. Thus, different conditions or patient groups have different microsystem designs. The various tasks of care are allocated to different members of a clinical team (including the patient), with the skill and training of each staff member matched to the work. Such organizations make thoughtful use of assistive personnel and alternative providers, and they ensure that each has the necessary resources by carefully designing the supply chain of equipment and information, simplifying workflow, and reducing work stress. They also harmonize the parts of their management system so that budgets, incentives, data, goals, clinical processes, educational programs, and team structures are all mutually reinforcing.4 Unit-level routines, such as joint ward rounds, team meetings, and executive “walk-arounds,” help tie microsystem components together.

    Attention to microsystem design and integration represents an important shift away from general-services-organization designs that use a single platform to meet the needs of many different patient groups and that focus on maximizing the use of scarce resources, such as operating-room slots, ICU beds, and physicians.

    The third habit is measurement and oversight. For many, measurement of clinical operations is driven by external audiences: payers, regulators, and rating agencies. Although high-value organizations share this reporting obligation, they primarily use measurement for internal process control and performance management. They collect more (and more detailed) measurements than those required for external reporting, selecting those that inform staff about clinical performance. For instance, of the 200-plus measurements used by Intermountain, more than two thirds were developed or refined internally rather than imported unmodified from external agencies. Moreover, such organizations integrate their measurement activities with other organizational priorities such as pay for performance, annual target setting, and improvement activities, making measurement an integral part of accountability and performance management. For example, each year Intermountain’s board selects a different group of measurements from the institution’s overall measurement set to use for annual quality and efficiency bonuses.