| Site Index |
|
|
         |
|
Geriatrics: ••Winter 2003/Vol. 7, No. 1 |
|
|
|
|
|
|
|
Care
of Common Medical Conditions in a Managed Care Program for Uninsured Adults.
| pdf
>> A
preliminary analysis of this material was published as an abstract as
follows: Swaney RE, Price DW, Goodspeed JR, Steiner JF. Quality of care
in a managed care program for the medically indigent. J Gen Intern Med
1999;14(suppl 2):124.
Background In 1998, more than 32 million adults in the United States were medically uninsured.1 The uninsured population has increased by an average of 1 million people per year since 1987.1 Most uninsured adults are from working families that do not qualify for Medicaid but cannot afford private health insurance.1 Although more than half of uninsured Americans are white, minority groups are disproportionately represented.1 Many trends contribute to the growing number of uninsured people, including rising costs of health insurance, workers shifting into low-wage or part-time positions where health benefits are not provided, and increased sharing of health care costs with employees.1 The majority of uninsured patients desire health coverage. In 1995, only 8% of the uninsured said they did not want or need health insurance; uninsured adults were four times more likely than insured adults to report that they did not receive medical care they believed was necessary.2 In 2000, more than a third of uninsured, nonelderly adults postponed medical care.1 In 1995, a quarter had not filled a prescription due to cost.2 Most uninsured adults have no consistent source of care; one in six consider an emergency department their regular source of care.1 The uninsured are much less likely to receive preventive services1 or early intervention services3 than their insured counterparts, and uninsured people are more likely to be hospitalized for "avoidable conditions"4,5 and to suffer adverse outcomes6-8 as a result of delayed care. Interest in managed care delivery systems for disadvantaged patients has been stimulated by these patients' need for improved access to preventive and early intervention services and by the increasing costs of care delivered in emergency departments and hospitals. As of 1993, 36 states either had or were developing managed care programs for their Medicaid recipients.9 As of June 30, 1992, 3.6 million Medicaid beneficiaries were enrolled in managed care plans.9 Evaluation of Medicaid managed care programs has demonstrated reduced use of subspecialists and emergency departments but little change in numbers of primary care visits, quality of care, or overall health care costs.10-22 Recent legislation has improved access to care for uninsured children by expanding Medicaid programs, developing new insurance mechanisms, and providing incentives to enroll children in private health care systems.23,24 However, few initiatives have been developed to improve access to care for uninsured adults. Kaiser Permanente (KP) of Colorado found that 326 previously uninsured children and adults who were enrolled in a program providing subsidized insurance coverage had similar medical utilization and moderately increased costs of care compared with an age- and sex-matched control group of commercially insured patients.25 To our knowledge, however, clinical process and outcome of care measures in other types of managed care programs for uninsured adults have not been systematically evaluated. In this study, we assessed clinical process and outcome of care measures in a primary-care-based, managed care pilot program for uninsured adults with one or more of three common chronic conditions in adults: asthma, diabetes mellitus, and hypertension. We hypothesize that this program would improve process and outcome of care measures for these patients through enhanced access to primary care services. Program Description The Colorado Resident Discount Program (CRDP) was implemented in 1983 to provide access to care for uninsured Colorado residents who fail to qualify for federal entitlement programs. The program was intended primarily to provide funds for emergency and hospital care; any remaining funds were to be used for outpatient services, including primary care. Historically, CRDP patients used emergency departments rather than primary care or preventive services for the majority of their care, a pattern that resulted in little continuity of care. At University Hospital (UH) in Denver, Colorado, some limited access to primary care for CRDP patients was provided in resident clinics. Direct referrals from the emergency department to subspecialty clinics were common, and little continuity or coordination of care existed. UH costs for CRDP patient care escalated rapidly, but there was little information on outcome of care for these patients. In 1992, UH, the School of Medicine at the University of Colorado Health Sciences Center (UCHSC), and KP of Colorado began developing a primary-care-based, managed care pilot program for CRDP patients who already received care at UH. The program, named CU CARE, began in June 1995. CU
CARE patients were mainly from the five suburban counties surrounding
Denver. Two family physicians, two general internists, and four nurse
practitioners provided care for almost 12,000 patients at the peak of
the program, until market and financial pressures forced a 50% decrease
in funding and rapid downsizing in December 1996. More than 90% of CU
CARE patients were age 19 through 64 years. Sixty-six percent were white,
57% were female, and 44% were single or never married. The program characteristics,
demographics, and chronic disease burden of the CU CARE population have
been described elsewhere.26,27
Figure
1. Patient selection process for establishing study cohorts Despite anticipated differences in income and medical comorbidity, Medicaid patients' access to UH primary care services was comparable to access by medically indigent patients, but their access was unaffected by the CU CARE program. We compared medical utilization, clinical process measures, and outcome of care measures for Medicaid (MC) patients to that for medically indigent patients who had made at least one visit to the CU CARE program in its first year (CUC). We used the same factors to compare care of CUC patients to the population of medically indigent (MI) patients who received care from other clinics within UH but who were not yet enrolled in the CUC program. The MI cohort included both patients who were not reached by CUC enrollment personnel and a few patients who declined to be enrolled. Study Conditions Patient Selection and
Study Groups We reviewed medical records for all eligible MC, CUC, or MI patients who had diabetes or asthma, all eligible CUC and MI patients who had hypertension, and a convenience sample of eligible MC patients who had hypertension. Patients with more than one of the index conditions were reviewed for each condition separately. Patients whose records showed no evidence of any index condition, who had a new diagnosis of the index condition during either study year, or whose medical records could not be retrieved were excluded. Of the 452 patients with hypertension, 66 MC patients were excluded because their medical records were not sampled, and 164 other patients were excluded due to inaccurate diagnosis or a missing chart. Overall, of the 3642 patients initially identified as potentially eligible for inclusion on the basis of billing data alone, chart reviews were completed for 76.9% of patients with asthma, 72.1% of patients with diabetes, and 49.1% of patients with hypertension. Using billing data for initial patient identification undoubtedly failed to detect some patients who had an index condition. However, the patient selection system we used identified cohorts of continuous users of the UH health care system, which was necessary for methodologically sound evaluation of the effects of the CU CARE program.
Selection
of Dependent Variables Statistical Analysis Descriptive statistics of demographic characteristics and outcomes were produced. Median and selected percentiles were reported for count data and nonnormally distributed continuous data. Mean and 95% confidence intervals were reported for normally distributed continuous data. Statistical significance was defined at p < 0.05. Proportions were reported for dichotomous variables. Baseline characteristics of patients--including age, gender, race, and comorbidity (presence of more than one index condition, complications from an index condition, or commonly occurring coexisting medical conditions)--were compared between payer cohorts using the Wilcoxon rank sum test and the chi-square test. Patient age, race, gender, and comorbidity were entered as covariates in all analyses. Because of patient-related clustering effect across years, mixed models were created for outcomes of interest. When a variable or its proper transformation was normally distributed (eg, weight and mean systolic blood pressure), a linear mixed model with different intercepts and slopes for payers was fitted.28,29 Change of an outcome (during the study years) within a group was tested for significance by examining whether the slope was zero, and the difference of change between groups was tested for significance by examining whether the difference in slopes between groups was zero. Nonlinear mixed models30,31 were fitted for count data (eg, number of clinic visits) and for dichotomous data (eg, eye examination). All analyses were conducted using SAS, Version 8.2 (SAS Institute, Inc, Cary, NC). Results Table 1 shows demographic and comorbidity data for the study groups. CUC diabetic patients were significantly older than both MI (p = 0.003) and MC (p = 0.02) diabetic patients. Significantly more MC (than CUC) patients with asthma were female (p = 0.01). Significantly fewer MC than CUC hypertensive patients were white (p = 0.03), and more MC than CUC hypertensive patients had one or more comorbid medical conditions (p = 0.004). Patients with Asthma Compared with MI patients, CUC patients showed trends toward increased primary care and decreased specialty care visits, although neither reached statistical significance. CUC patients were more likely than MI patients to do peak flow monitoring at home (p = 0.012). Patients with Diabetes Compared with MI patients with diabetes, CUC patients with diabetes had significantly increased numbers of outpatient (p < 0.001), primary care (p < 0.001), specialty, (p = 0.042) and nurse (p = 0.0014) visits. CUC patients with diabetes were also more likely to have glucose readings recorded in their charts (p = 0.053). CUC patients with both hypertension and diabetes were more likely to be taking angiotensin-converting enzyme inhibitors (p = 0.01) than MI patients with hypertension and diabetes. CUC patients with diabetes had greater improvement in HgA1c values in the program year than did MC patients (p = 0.052) or MI patients (p = 0.0047). Patients with Hypertension Results Across Cohorts
Table 5 shows improvements in 7 of 16 utilization measures for CUC patients compared with MC patients, and 9 of 16 utilization measures for CUC patients compared with MI patients. However, only a few improvements in clinical process or outcome of care measures for CUC patients compared with the MC or MI cohorts (4 of 27 clinical process or outcome measures in each comparison). In one measure--percentage of hypertensive patients with diastolic blood pressure <90 mm Hg--the MC cohort did better than the CUC cohort (p = 0.019). Discussion In this evaluation of a primary-care-based, managed care pilot program for uninsured adults, CU CARE, we found a short-term increase in utilization of primary care services for enrolled patients who had asthma, diabetes mellitus, or hypertension. Clinical process or outcome-of-care measures for CU CARE patients improved only marginally compared with patients receiving Medicaid or compared with uninsured patients who were not enrolled in CU CARE. These results are consistent with published analyses of Medicaid managed care programs, where short-term effects on quality of care were modest.10-22 However, several findings deserve further comment. The increase in visits by CUC patients with hypertension may have been driven both by decreased blood pressure control (the percentage of these patients with mean systolic BP <140 mm Hg and mean diastolic BP<90 mm Hg decreased during the study) as well as by increased access to care. Improved glycosylated hemoglobin levels for the CU CARE diabetic patient cohort may result from regression to the mean, and we cannot exclude the possibility that the patients in the three diabetes cohorts differed in duration of illness. The small number in the MI cohort of patients with asthma (n = 17) may have limited the ability of the analysis to detect additional differences between CU CARE and MI asthmatic patients. The percentage of patients with more than one emergency department visit or hospital admission was not significantly different between CUC and MI cohorts or between CUC and MC cohorts of patients who had diabetes or asthma. However, in a different study, we found overall significantly decreased emergency department visits and hospital admissions in the entire CU CARE cohort compared with a matched MC cohort and a matched cohort of "self-pay" patients who were not enrolled in the Colorado Resident Discount Program for medically indigent patients.32 Perhaps the most important finding of our study was the difficulty we experienced in identifying a stable population of patients who had no private health insurance and who could be used to assess effectiveness of the CU CARE program. As shown in Figure 1, only 17.1% of patients with a billing claim for asthma, diabetes, or hypertension used the UH system and maintained the same payer over the entire 30-month study period.
Why did the Program Fail
to Show Substantial Improvement in Care? Despite the sliding copayment schedule, many patients (particularly those with higher copayments) may not have been able to afford prescription medications, a barrier which could have reduced their adherence to a medication regimen. In addition, affecting patient lifestyle takes time, and a lag time from lifestyle changes to improved outcome occurs. The period of analysis was probably too short to reflect an effect from change in patient behavior. Limitations One important limitation of our study is the inability to generalize our findings to the much larger population of patients without private health insurance and who have one of the study conditions. Only a small proportion of the initial group of patients identified who had these common health conditions was eligible for record review, because we required a UH visit for the study condition in each year and a consistent payer class to establish study eligibility. The same criteria necessary for methodologically sound program evaluation proved hard to satisfy for this population, who often seek care sporadically or from multiple sites. As a result, the small sample sizes limited the statistical power of the study to detect differences in process or outcome of care measures. The results may also not be generalizable to other medically indigent populations, which may differ from our study population in demographics, prevalence or severity of illness, local system of health care for the uninsured, or primary care provider mix (eg, family practice or internal medicine; physician or nurse practitioner). The analysis
was retrospective, and patients were not randomized into different cohorts.
Inclusion in the CUC cohort may have been subject to selection bias,
because indigent patients visiting the emergency department were "targeted"
for enrollment into the CU CARE program at the time of their visit.
The MI cohort included patients who were not reached by CUC enrollment
personnel as well as a few who declined to be enrolled in CU CARE. This
distinction is important, because patients who declined enrollment may
have differed in health status or views on health care from patients
who were not reached. Table 1 shows some demographic and clinical differences
between CUC enrollees and the control group of MC patients. Unmeasured
differences may well have existed between groups as well. For example,
CUC patients may have sought care at other institutions, thus making
it difficult to control for the potentially confounding influence of
lack of continuity of care. Because claims data were used to identify
patients who had the study conditions, some patients who had one or
more of these conditions but who did not have an ICD-9coded visit
for that condition would have been excluded from the study. However,
lack of feasibility of a randomized trial in this setting realistically
reflects the constraints in assessing program effectiveness in practice.
We therefore believe our findings are relevant to policymakers who are
considering systems of care for the medically uninsured. The analysis did reveal some statistically significant improvement in process and outcome of care measures, although because of multiple comparisons, some of the improvement could possibly have occurred by chance alone. However, the purpose of this analysis was to look for trends in care of adults with chronic conditions, and no short-term systematic trends toward improved process or outcome measures were evident for any of the three health conditions studied. The possibility still exists that with a longer period of analysis, a larger sample size, and a more mature delivery system, significant improvements may be realized. Conclusions Table 6 summarizes the important conclusions from this study. Our evaluation showed that the CU CARE program increased utilization of primary care services by uninsured adults who maintained continuity of care within the UH system. However, CU CARE did not lead to consistent short-term improvement in process or outcome of care measures for patients with three common medical conditions. CU CARE program personnel were unable to reach most of the people who may have benefited from enrollment because of high mobility and frequent transition of insurance status in the uninsured population. We conclude that increases in primary care utilization alone do not automatically lead to short-term improvement in quality of care for medically indigent populations, which is consistent with the results of quality-of-care assessment in Medicaid managed care programs.14-16 Programs providing primary care to uninsured populations should recognize that their efforts to improve process and outcome of care measures may need to be continued for prolonged periods to demonstrate benefit. Such programs should also consider strategies for reaching beyond the small proportion of patients who have one or more specific health conditions and who maintain continuity of care into the larger population of people who have the health condition or conditions of interest. In fact, high mobility of the patient populations that programs such as the CU CARE program attempt to serve, lack of continuity of care, and inability to enroll many eligible patients remain fundamental obstacles to improving quality of care for these populations.
Acknowledgments
References
To Original Research contents list >>
|
|
|
|
|
|
|
|
|
|
|
 |
|
