External Incentives, Information Technology, and Organized Processes to Improve Health Care Quality for Patients With Chronic Diseases FREE

Recent reports, including 2 by the Institute of Medicine (IOM) of the National Academy of Sciences, argue that the quality of health care in the United States falls far short of biomedical knowledge and that this gap in quality is primarily a failure of organization, rather than of individual physicians.^1- 7 The IOM and others have called for the implementation of organized processes to improve quality and have argued that government and large private purchasers of health care should provide physician organizations (POs) with incentives to implement such processes.^8- 14 The IOM also has advocated government financial assistance to POs to improve their clinical information technology (IT), which is considered fundamental to organized attempts to improve quality of care.^2,15- 16

Despite this attention and despite a growing body of research supporting the effectiveness of organized processes in improving quality of care,¹⁷ little information is available to answer 4 fundamental questions: (1) To what extent do POs—medical groups and independent practice associations (IPAs)—currently use organized processes to improve quality?¹⁸ (2) Do POs have external incentives to improve quality? (3) What clinical IT capabilities do POs have? and (4) Are external incentives and clinical IT capabilities associated with increased use of organized processes by POs to improve the quality of care?

To address these questions, we conducted a National Survey of Physician Organizations and the Management of Chronic Illness (NSPO). We used 5 separate databases to develop the most extensive census available to date of medical groups and IPAs with 20 or more physicians, and we conducted a telephone survey with 1040 of these POs. We focused on POs with 20 or more physicians because, compared with individual and small group practices, these larger organizations should have more capacity to invest in the management time and expertise, clinical IT systems, and skilled nonphysician staff that are thought to be necessary for the creation and maintenance of organized processes to improve the quality of care.^19- 22 We called these organized processes care management processes (CMPs) and hypothesized that the extent to which a PO uses CMPs will vary with its IT capability and with the incentives for improvement of quality provided by health insurance plans and health care purchasers.

The NSPO survey focused on 4 types of CMPs— case management, performance feedback to individual physicians, use of disease registries, that is, lists of a PO's patients with a particular chronic illness that make it possible for the PO to organize care for patients with that particular illness, and use of clinical practice guidelines in conjunction with physician education and electronic or chart-based reminder systems. Support for self-management skills for chronically ill patients was not assessed for the 4 individual chronic diseases surveyed but only for managing their disease in general (Box).

Case Management
Case management is available at the request of the physician and/or assigned to all severely ill patients with the disease.

Physician Feedback
Physicians receive feedback on specific practices (eg, use of anti-inflammatory medication in asthma, angiotensin-converting enzyme inhibitors in congestive heart failure, retinal screening in diabetes). Depression was omitted for this category.

Disease Registry
This is maintained for patients with the disease.

Clinical Practice Guidelines
These are (1) adopted and (2) physicians are trained in the guidelines, and (3) the guidelines must be present in patient charts or in reminder systems or in order-entry systems. An answer of yes for use of all 3 is required.

Self-management Skills
Programs to teach chronically ill patients skills for better managing their illness in general, not for each of the 4 chronic diseases surveyed.

*For each care management process, the organization receives 1 point for each disease for each affirmative answer (except for feedback to physicians for depression) and 1 point for use of programs to teach patients self-management skills.

We chose these CMPs because evidence is growing that they are effective, individually and especially in combination, in improving quality of care.^17,23- 27

Case management programs repeatedly have been shown to improve outcomes for patients with congestive heart failure (CHF), diabetes, and mixed comorbidities.²⁸ While clinical practice guidelines alone have not been demonstrated to change physician performance,²⁹ a Cochrane review and recently published studies, including a meta-analysis, have found that guidelines in conjunction with physician education or reminder systems do improve physician management of several chronic conditions and may improve clinical outcomes.^17,30- 31 Performance feedback to physicians has been shown to improve medical practice in both a Cochrane review and a separate meta-analysis, although the effect was small.^17,32 Disease registries have been shown to be important components of health care management.³³ Patient self-management education can succeed in improving clinical outcomes according to a Cochrane review of use of this technique for patients with asthma³⁴ and according to randomized controlled trials of patients with diabetes^35- 36 and with a variety of other chronic diseases.^37- 38

The NSPO survey focused on POs' use of CMPs for asthma,^39- 40 CHF,^41- 43 depression,^44- 48 and diabetes.^26,49- 50 These 4 diseases are directly responsible for approximately 140 000 deaths each year in the United States,⁵¹ generate at least $173 billion in annual costs,^52- 55 and are conditions for which evidence suggests that CMPs can improve the quality of health care.^17,23,56

To generate as complete a national census as possible of medical groups and IPAs with 20 or more physicians, we combined databases from the Medical Group Management Association, Englewood, Colo; Dorland Healthcare Information, Philadelphia, Pa; National IPA Coalition, Oakland, Calif; American Hospital Association, Chicago, Ill; and Virginia Commonwealth University Study of Physician Organizations, Richmond (this database merges data from 6 sources, including a survey of medical groups by the American Medical Association, Chicago, Ill) (R.R.G., unpublished data, 2002). The National Opinion Research Center at the University of Chicago made preliminary telephone calls to all 3233 organizations identified to verify their type and size and to arrange an appointment for a telephone survey. On the basis of these calls, 1646 organizations (51%) were excluded because they were out of business or could not be located (21%), were not a medical group or IPA (20%), had fewer than 20 physicians (17%), were duplicates (20%), or were excluded for miscellaneous reasons (11%). In addition, we excluded the 10% of POs that were composed of radiologists, pathologists, chiropractors, podiatrists, ophthalmologists, anesthesiologists, or emergency department physicians.

Of the 1587 organizations remaining, 1104 (70%) responded to the survey. Information on nonrespondents is limited: they did not differ from respondents by size or by state where they were located, but response rates were significantly higher for IPAs (79%) than for medical groups (66%; P<.001). Of the 1104 respondents, we excluded 64 organizations that stated that they did not treat any of the 4 chronic diseases we surveyed, leaving 1040 POs (693 medical groups and 347 IPAs) included in our study.

The NSPO survey was developed based on review of the literature on health care quality improvement, on dimensions of the Assessment of Chronic Illness Care model,⁵⁷ on feedback from the NSPO's national advisory committee, and on input from a focus group of 9 medical directors from POs. The survey was revised after a pilot test was conducted with 36 medical groups and 10 IPAs. The survey instrument is available from the corresponding author.

Organizations that agreed to participate were sent in advance a packet with a worksheet to be completed prior to the interview. Specially trained NORC interviewers then conducted 60-minute telephone surveys between September 2000 and September 2001 with each organization's president, chief executive officer, or medical director, using Computer Assisted Telephone Interviewing.

To assess POs' use of CMPs, interviewers asked, for each of the 4 chronic diseases, whether the PO used case management, a disease registry, clinical guidelines, and feedback to physicians (feedback was not asked for depression) (Box). Organizations also were asked whether they have general programs to teach patients self-management skills for chronic diseases. To avoid positive but overly general answers that might not accurately reflect an organization's use of CMPs (eg, "Yes, we have guidelines"), survey questions were designed to probe for more specific answers (eg, "Are the guidelines placed in patient charts, reminder systems, or order entry systems?" and "Do you feed back data to physicians on appropriate use of ACE [angiotensin-converting enzyme] inhibitors for patients with CHF?").

To assess the external incentives for a PO to improve quality of care, the survey asked about a range of 7 incentives: whether the PO received a bonus from health plans (ie, additional income from health plans for scoring well on quality measures), public recognition (eg, through report cards), or better contracts with health plans for quality and whether it was required to report Health Plan Employer Data and Information Set data,⁵⁸ outcomes data, results of quality improvement projects, or patient satisfaction data to an outside organization (Table 1).

The use of clinical IT systems⁵⁹ was assessed by asking POs whether they use an electronic data system and, if so, whether the IT system includes a standardized problem list, the majority of physician progress notes, medications prescribed, medication decision support (ie, reminders and/or drug interactions information), laboratory results, and radiology results (Table 2).

We also assessed other organizational characteristics of POs that may affect the use of CMPs: the survey provided demographic information on the type of PO (medical group or IPA), number of physicians, practice type, practice ownership, and US location (divided by the American Medical Association census regions) (Table 3).

Health maintenance organizations (HMOs) also might affect POs' use of CMPs. We used InterStudy data to define HMO penetration as the number of people enrolled in HMOs in a PO's county divided by the county population.⁶⁰ The survey asked whether a PO took at least some risk for hospitalized HMO patients (964 POs answered this question) and whether HMOs delegate utilization management responsibility for hospitalized HMO patients to the PO (965 POs answered this question) (Table 3).⁶¹

We used multivariate linear regression to measure the association of external incentives, IT capability, characteristics of POs, HMO penetration, and physician group health plan contracting characteristics with POs' use of CMPs. The outcome variable in our main model is a summary measure—a Physician Organization Care Management Index (POCMI)—with a range of 0 to 16. An organization's POCMI score is the number of CMPs used by the PO for the 4 chronic diseases: 1 point for each of the 4 CMPs for each disease (with the exception of feedback to physicians for depression, which was not asked) plus 1 point for use of programs to teach patient self-management support skills for chronic illnesses in general (Box). Omission of the feedback question for depression meant that depression was not weighted as heavily (3 possible points for CMP use for depression) as the other diseases (each of which has 4 possible points for CMP use). Regression results did not change significantly when depression was excluded from the model nor when depression was weighted equally to the other diseases.

The POCMI scores were normally distributed. The POCMI index exhibited a high level of internal consistency reliability with a Cronbach α of 0.86. Analyses were performed using SAS version 8.2 (SAS Institute Inc, Cary, NC).

The POCMI index treats the use of each type of CMP as equally important. Some CMPs may be more effective in improving quality of health care than others, but we did not believe that sufficient evidence existed to weight them differentially in our analysis.¹⁷ We did, however, run regression analyses using a variety of different weightings; the results of these analyses did not differ significantly from our main model.

The 4 external incentives concerning reports to an outside organization (of Health Plan Employer Data and Information Set data, outcomes data, results of quality improvement projects, and patient satisfaction data) are highly correlated with each other and are included in our main regression analysis as a single summary variable with a range of 0 to 4. The 3 other incentives are included individually. Results did not change when all 7 external incentives were included individually, except that the reporting requirements variables, which are significant as a group, are not significant individually. We included the 6 IT measures as a single index variable, with a range of 0 to 6, in our main model. Results did not differ when the IT measures were examined individually rather than as an index in a regression analysis.

Ninety-three POs (9%) stated that they treated only 3 of the 4 diseases, and 83 (8%) stated that they treated 2 or fewer diseases. In our main model, we included these POs and assigned them scores for each disease they did not treat equal to the mean score for that disease for all POs that treated that disease. We also conducted regression analyses that included only the 947 POs that treated at least 3 of the 4 diseases. Results from these analyses did not differ significantly from our main model.

The mean POCMI score was 5.1, indicating that the average PO used 32% of the 16 CMPs (Table 4); 15.2% used none, 49.6% used 4 or fewer, and 5.3% used 13 or more CMPs. Case management was the most commonly used CMP: that is, the mean percentage of POs using case management across all 4 diseases was 37.6% (Table 5). Clinical guidelines used in conjunction with reminders placed in charts or in the electronic medical record were least commonly used (28.5%) (Table 5). Care management processes were implemented most often for diabetes (42.5% of POs) and least often for depression (17.1% of POs). Few POs used any one type of CMP across all 4 illnesses (range, 10.5%-18.5%, for clinical guidelines, disease registry, physician feedback, and case management), and fewer (3.2%-12.7%, for depression, asthma, CHF, and diabetes) used all 4 types of CMP to focus on any 1 illness (Table 5). Self-management programs for chronically ill patients were provided by 589 (56.6%) of POs (not included in Table 5 because the survey asked about self-management programs in general, not for each of the 4 diseases).

The results of multivariate analysis showed that the presence of external incentives to improve quality of care was most strongly associated with use of CMPs (Table 6). When all other variables, including other external incentives, are held constant, POs that received public recognition for scoring well on quality of care measures used 1.3 more CMPs (P<.001). Physician organizations that reported receiving better contracts for scoring well used 0.74 more CMPs (P = .007). Requiring POs to report quality of care data and activities to outside organizations also was significantly associated with increased CMP use (P<.001). Receiving a bonus for scoring well on quality of care measures was not significantly associated with CMP use (P = .08), possibly because POs generally reported that the amounts received were very small.

The mean number of external incentives reported by POs was only 1.7 of the 7 external incentives surveyed (Table 1). Three hundred forty-one (32.8%) POs had no incentives, 776 (73.6%) had 2 or fewer incentives, while only 177 (17.0%) had 4 or more incentives to improve quality (data not shown).

Clinical IT also is significantly associated with CMP use: each additional capability is associated with the use of 0.37 additional CMPs (P<.001) (Table 6). However, the average PO had only 1.4 (23%) of the 6 clinical IT capabilities surveyed (Table 2). Five hundred nineteen (49.9%) of POs had none and 816 (78.5%) had 2 or fewer (data not shown) of the 6 clinical IT capabilities surveyed.

The strong association of external incentives and IT with the use of CMPs was found in every model we tested, including the POCMI model presented in this article, which evaluates the association with all types of CMPs surveyed as a group, the POCMI model presented in this article with depression excluded, the POCMI model with a variety of weights assigned to individual types of CMPs (data not shown), the POCMI model run for each of the 4 chronic diseases individually (data not shown), and models in which the association is tested for each type of CMP separately (data not shown).

Health maintenance organization penetration, PO size, hospital or health plan ownership of a PO, and the percentage of patients for which a PO is delegated utilization management also are statistically significantly associated (P<.05) with the use of CMPs by POs, but their impact is small (Table 6). For example, an increase of 100 physicians in a medical group is associated with use of only one fifth of an additional CMP, and an increase from 0 to 50% in the percentage of patients for whom the PO is delegated utilization management is associated with use of only 1 additional CMP.

Many medical groups and IPAs have implemented CMPs to improve the quality of care for their patients. But on average, POs use 5.1 CMPs, which is less than one third of the 16 CMPs surveyed. These findings can be interpreted as encouraging that POs are using any CMPs at all. However, the fact that 50% of POs use 4 or fewer of the 16 CMPs and that POs lacked registries of patients with a chronic disease 69% of the time suggests that organized processes to improve the quality of health care remain relatively uncommon in POs. It is difficult for an organization to provide proactive health care for a specific group of patients if the organization lacks these registries, that is, it does not know who these patients are.

Our findings support the IOM and others^8- 14 who argue that rewarding POs for improving quality of care is important. Giving POs public recognition and better contracts for scoring well on quality measures is associated with the use of 2.0 additional CMPs—a substantial increase, given that the average PO uses only a total of 5.1 CMPs. However, our data also confirm the widespread perception that incentives for POs to improve quality are uncommon^10,20,62: 32% of organizations reported having no incentives and 74% had 2 or fewer of 7 incentives surveyed to improve quality of care.

Our findings also support the belief that clinical IT is important, but that most POs lack clinical IT capabilities.^10,20 Each additional IT capability was associated with the use of 0.37 more CMPs, but 50% of POs had none of and more than 78% had 2 or fewer of the 6 IT capabilities surveyed.

The NSPO is the first study, to our knowledge, to provide national data on the use of CMPs by POs to improve the quality of health care and to show that clinical IT capabilities and external incentives to improve quality are each strongly associated with POs' use of CMPs. This strong association was present in every model for which we conducted statistical analyses, including models using a variety of weights for individual types of CMP, models testing the association for individual types of CMP, and models testing the association for individual diseases.

Our study has limitations, each of which suggests areas for future research. First, although respondents were assured that neither they nor their organizations would be identified, some may have overstated their organization's CMP use, even though the survey questions were designed to minimize this possibility. Also, it is possible that there may have been a response bias, such that organizations that use few or no CMPs were less likely to agree to be surveyed. For both these reasons, our data may somewhat overestimate the extent of CMP use among POs in the United States. Second, even though POs were asked what CMPs were used by their organization as a whole, we do not know whether each individual practice site within a PO was in fact using the CMPs claimed for the organization. It is possible that some sites may have been using less and some using more CMPs. Third, in this cross-sectional study we demonstrate strong associations between certain variables and CMP use, but we cannot show the direction of causality, which may work to varying extents in both directions. For example, we suggest that it is plausible that the presence of external incentives to improve quality would induce POs to implement CMPs, but we cannot exclude the possibility that health plans and purchasers are more likely to provide such incentives to POs that already have some CMPs in place, that is, that have shown some capability to manage health care. Analogously, clinical IT increases organizations' ability to implement CMPs, but organizations planning to implement CMPs may be more likely to invest in clinical IT. Simply providing clinical IT systems (eg, through government financial support) to POs that lack both interest and incentives to improve quality may be unlikely to have much effect.

Underlying our study is the hypothesis that CMPs improve the quality of health care. Although evidence to support this hypothesis is accumulating, more research is needed to further explore the relationship between CMP use and clinical outcomes, specifically whether certain individual CMPs and combinations of CMPs are more effective than others for specific chronic diseases. This research will be difficult.^61,63 Our study also assumes that organizations of 20 or more physicians are more likely to implement CMPs than smaller organizations, but the extent to which physicians who work in smaller practices use CMPs is unknown. Further research is needed to identify the extent of CMP use and its relationship to clinical outcomes as well as the factors associated with CMP use in these practices. Finally, we are able to explain 28% of the variance in CMP use among POs, but the literature suggests that factors not evaluated in this study, particularly an organization's leadership and culture, also are important and should be a subject for further research.^64- 68

The findings from the NSPO have a number of important implications. For medicine as a profession, the fact that the use of CMPs is relatively uncommon in POs raises the question of whether such processes will be developed primarily by other entities, such as health plans and pharmaceutical companies, without adequate coordination with physicians. For government and large private purchasers of health care, this study provides strong confirmation for the IOM's assertions that organized processes to improve quality of care are not common in POs and that many physicians work in practices that have neither incentives nor IT capabilities to improve quality. Our data support policy recommendations by the IOM and others that POs should be given incentives to improve quality and support for developing clinical IT systems. Attempts to provide incentives for quality of care, though still exceptional, are becoming more common.^69- 71 Such incentives can complement the efforts of organizations, such as the Institute for Healthcare Improvement, Boston, Mass,^72- 73 and the MacColl Institute for Healthcare Innovation, Seattle, Wash,^74- 75 that are working to help POs develop organized processes to improve the quality of health care.