+ Models
1
Diabetes Research and Clinical Practice xxx (2005) xxx–xxx www.elsevier.com/locate/diabres
2
4
Geographic variation, physician characteristics, and diabetes care disparities in a metropolitan area, 2003–2004 Jeff McGinn a,*, Cathy Davis b,1
5 6 7
OF
3
a
b
8 10 9 11
Primaris, 200 North Keene Street, Columbia, MO 65203, USA UAW Ford Community Health Care Initiative, 6000 N. Oak, Suite 300, Kansas City, MO 64118, USA Received 12 May 2005; received in revised form 3 October 2005; accepted 12 October 2005
29
Abstract
DP
Quality improvement begins with quality measurement. Quality improvement continues by examining variation in patient standards of care. Diabetes is a complex chronic disease requiring aggressive care to improve and prolong life. Identifying and determining what factors explain care variation—including physician office location, physician characteristics, practice factors, and the translation of knowledge over time—is an area of important research. Determining physician practice characteristics’ impact on these diabetes measures is an important step in understanding diabetes care in greater Kansas City. Aggregated diabetes HEDIS data from numerous private insurance plans from greater Kansas City, and associated practice factors from these same providers serves as a large and representative source of information to evaluate the impact of these factors on diabetes care. Using both multivariate and logistic methods, we find that the variation in care is largely explained by physician office location, but there is statistical explanatory significance for physician age on A1c testing rates. Also, nephropathy screening rates are positively related to whether a physician is a member of a group or a solo practice. The location of a practice has significant effects for diabetes care because physician office location approximates to some extent patient characteristics. This is not to say that physician practice factors are unimportant, rather that diabetes care is complex, and requires a fruitful interaction between physician and patient. Quantifying the interaction between an informed, activated patient and a prepared, proactive physician is difficult to observe and measure, and evaluating testable hypotheses about this interaction is correspondingly difficult. # 2005 Published by Elsevier Ireland Ltd.
EC TE
13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
RO
12
Keywords: Geographic variation; Physician characteristics; Diabetes care disparities
30 31
1. Background
RR
32
37
translation of diabetes care knowledge—is an area of increasing research [6]. The Kansas City Quality Improvement Consortium (KCQIC) contracted with Primaris, a healthcare consulting company, to create physician-specific dashboards combining HEDIS diabetes measures from the Aetna, Blue Cross/Blue Shield (BCBS) of Kansas City, CIGNA, Coventry, FHP, FirstGuard, Humana, and United Health Care (UHC) insurance plans. Included physicians were from five counties in and around Kansas City: Jackson (MO), Platte (MO), Clay (MO), Johnson (KS), and Wyandotte (KS). These dashboards detail individual physician results, the results of all included physicians,
Quality improvement begins with quality measurement, and continues by examining variation in patient 35 standards of care. Identifying and determining what 36 factors explain care variation—including geography, 37 physician characteristics, practice factors, and the 33
CO
34
UN
* Corresponding author. Tel.: +1 573 817 8300x143; fax: +1 573 817 8330. E-mail addresses:
[email protected] (J. McGinn),
[email protected] (C. Davis). 1 Tel.: +1 816 453 4424; fax: +1 816 453 4107.
0168-8227/$ – see front matter # 2005 Published by Elsevier Ireland Ltd. doi:10.1016/j.diabres.2005.10.023
DIAB 3469 1–8
38 39 40 41 42 43 44 45 46 47 48 49 50
+ Models
2
J. McGinn, C. Davis / Diabetes Research and Clinical Practice xxx (2005) xxx–xxx
Table 1 Socio-economic characteristics of the Kansas City Area, by county Countya
Clay Jackson Johnson Platte Wyandotte
Hispanic (%)
Median household income, 2002
2.7 23.3 2.6 3.5 28.3
3.6 5.4 4 3.0 15.9
$52,169 $41,833 $66,974 $59,141 $32,465
Percentage of pop. below poverty line (%)
Patients per county
Physicians per county
5.2 11.3 4.3 5 15.4
4186 13491 699 2003 3693
179 808 84 91 223
United States Census Data for race and income can be found at: http://www.census.gov/cgi-bin/saipe.
OF
a
Black (%)
50
54 55 56 57 58 59 60 61
and extremities [1]. Due to the complexity of managing diabetes and its complications, aggressive disease management is critical. Blood chemistry, especially blood glucose and lipid control, is fundamental to the management of diabetes and its complications. The United Kingdom Prospective Diabetes Study (UKPDS) [1] showed that improved blood glucose level control is associated with sustained decreased rates of serious diabetes-related complications, including heart disease, blindness, and kidney failure [2]. The HbA1c test provides a ‘‘big picture’’ of diabetes blood glucose level control over.
1. the percentage of administered A1c tests for patients with diabetes; 2. the percentage of patients with A1c results greater than 9.5% in 2003, and greater than 9% in 2004; 3. the percentage of administered LDL-C tests for patients with diabetes; 4. the percentage of patients with LDL-C results greater than 130 mg/dl; 5. the percentage of retinal eye exam referrals for patients with diabetes; and 6. the percentage of performed nephropathy screenings for patients with diabetes.
3. 2004 Geographical factors
66 68 67 70 69 68 71 69 73 72 70 74 71 76 75 72 77 73 79 78 74 80 75 81 76 82 77 83 78 84 79 85 80 86 87 88 89 90
RR
65 67 66
Determining physician office location and other physician practice characteristics’ impact on these diabetes measures is an important step in understanding diabetes care in the greater Kansas City area. Physician office location matters because there are significant differences in the populations of these counties. Table 1 lists some notable differences from 2002 Census data, as well as physician and patient densities per county.
UN
2. Diabetes and its complications 91
CO
64 65
Diabetes can lead to a number of life-threatening complications that affect the eyes, kidneys, nerves, skin,
94 95 96 97 98 99 100 101 102 103 104 105
106
The county in which a physician or practice office is located has significant effects on a number of the diabetes measures. Jackson County represents 54.5% of all patients in the five-county area. A1c and LDL-C testing rates in Wyandotte County are statistically lower than the combined county rates. The rate of LDL-C testing in Jackson County is statistically less than the combined rate. In Platte County the retinal exam rate is significantly lower than the combined rate. Clay, Johnson, and the previous three months, and also indicates how well a person with diabetes is managing the disease. These tests allow patients to reduce their risk of diabetes-related complications such as cardiovascular disease, foot amputation, kidney failure, and blindness. For every 1% decrease in HbA1C level, there is a 38% reduction in a person’s risk for developing micro-vascular diabetes-related complications [3,4]. Also, an estimated 30–70% of renal failure could be prevented through the detection and treatment of early diabetic kidney disease, through procedures like nephropathy monitoring [4]. These and other clinical measures are reported through the Quality Control dashboard.
EC TE
62 64 63
RO
52 53
93
and the physicians in the top performing benchmark of physicians (Attachment 1). The Benchmark aggregates the top-performing physicians or practices in the upper 10th percentile of all practices for a given diabetes measure, provided the practice serves five or more patients with diabetes. Top performance calculations used the Achievable Benchmarks of Care (ABC) metric described in Keating et al., ‘‘Identifying achievable benchmarks of care: concepts and methodology’’ [5]. The diabetes measures are the following HEDIS indicators:
DP
51
DIAB 3469 1–8
107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128
+ Models
J. McGinn, C. Davis / Diabetes Research and Clinical Practice xxx (2005) xxx–xxx
3
Table 2 County level diabetes care measures, 2004a Clay (n = 2973), rate (%)
Jackson (n = 9426), rate (%)
Johnson (n = 481), rate (%)
Platte (n = 1498), rate (%)
Wyandotte (n = 2598), rate (%)
Combined counties rate (%)
Benchmark rateb (%)
A1c tested A1c poorly controlled (>9%)b,c
75.0 9.4
72.3 10.2
80.0 2.5
74.8 10.1
60.4 33.0
71 13
94 1
A1c results by ranged Under 7% 7–
