Association between NCQA Patient-Centered Medical Home Recognition for Primary Care Practices and Quality of Care for Children with Disabilities and Special Health Care Needs. RESULTS

01/01/2014

Sample Characteristics

There were 241,997, 28,797, and 389,932 CSHCN in FFS Medicaid and not living in institutional settings for more than 90 days in Louisiana, New Hampshire, and Texas, respectively, in 2010. Based on our measure for special health care needs, CSHCN comprised approximately 30 percent of the base Medicaid populations in all three states. Among CSHCN (children with a disability and/or a CDPS diagnosis) the proportion that had a disability varied widely across states -- ranging from 1 percent in New Hampshire, to 18 percent in Louisiana, to 31 percent in Texas. The variation in the proportion with at least one CDPS diagnosis was much narrower (90-100 percent). Variation in the percent of children with a disability in our samples across the three states is consistent with variation observed in the percent of children with disabilities in the overall Medicaid FFS population in each state (Appendix B).

Table 1 shows demographic characteristics, months of Medicaid enrollment, health status and zip code-level SES characteristics of treatment and comparison groups in Louisiana, New Hampshire and Texas. In Louisiana and Texas in the non-matched, "late recognition" analyses, there were more treatment group children in the younger age categories and more males, whereas in New Hampshire the overall age distribution between treatment and comparison was more balanced, although there were also more males in the treatment group. Also in the "late recognition" comparison group analyses in Texas, there were fewer children with Medicaid eligibility based on disabilities in the treatment group relative to the comparison group. Health status based on CDPS condition categories was generally similar for the "late recognition" comparison group analyses in all three states. Treatment group children in "late recognition" analyses in Louisiana lived in neighborhoods with higher percentage of Hispanic/Latino, White and other race residents compared to comparison group children. Treatment group children in Louisiana also lived in neighborhoods with lower rate of poverty and higher share or residents working full-time. SES characteristics were generally balanced in New Hampshire and Texas in "late recognition" analyses, although treatment group children lived in neighborhoods with a higher share of Hispanic residents and lower share of White residents in New Hampshire and treatment children in Texas resided in neighborhoods with higher share of White residents and lower share of Hispanic residents relative to the comparison group.

By design, the matched comparison groups in the three states were balanced on all demographic and health status variables (Table 1). Across zip code-level SES-related characteristics, treatment children in matched analyses in Louisiana lived in zip codes with a smaller share of White, non-Hispanic residents and a larger share of Black, non-Hispanic residents compared to the comparison group. Treatment group children in Texas lived in neighborhoods with a smaller fraction of Hispanic residents than the comparison group. Treatment group children also lived in neighborhoods with more women with college degrees or higher compared to children in the matched comparison group.

Most treatment providers in Louisiana had Level 1 recognition (79 percent), on a scale from 1-3 where practices with Level 1 recognition had fewer characteristics of a PCMH compared to practices at Level 2 or 3. Most of the "late recognition" comparison group providers in Louisiana (i.e., recognized in 2011) had Level 3 recognition (59 percent). All NCQA-recognized treatment and "late recognition" comparison providers in New Hampshire had Level 3 recognition. In Texas, all treatment providers had Level 3 recognition, but only a bare majority (56 percent) of comparison providers did.

FIGURE 1. Louisiana "Late Recognition" and Matched Comparison Group Analyses: Odds Ratios and 95% Confidence Intervals for Treatment vs. Comparison Group from Logistic Regression Models for Service Utilization and Care Coordinationa
See ALT TEXT at end of this figure.
NOTE:
  1. Models for all outcomes adjusted for age (0-1, 2-5, 6-12 or 13-18 years), sex, disability status (disabled vs. non-disabled), number of months enrolled in Medicaid, number of CDPS chronic conditions, number of CDPS prescription drug categories, skeletal and GI CDPS condition categories, CDPS prescription drug categories related to asthma/COPD, ADHD, inflammatory/autoimmune, and nausea), combined categories for CDPS condition and prescription drug categories (diagnoses and prescriptions related to cardiovascular disease, diagnoses related to psychiatric conditions or prescriptions for depression/anxiety or psychotic illness/bipolar, diagnoses for CNS conditions or prescriptions for multiple sclerosis/paralysis, diagnoses related to pulmonary disease or prescriptions for cystic fibrosis, diagnoses related to hematological conditions or prescriptions related to hemophilia/von Willbrands disease, diagnoses or prescriptions related to infectious diseases, including HIV/AIDs or prescriptions for HIV/AIDS, as well as diagnoses or prescriptions related to diabetes, and diagnoses or prescriptions related to renal disease), as well as a series of zip code-level variables from the ACS (percent Hispanic/Latino, non-Hispanic/Latino White, non-Hispanic/Latino Black, non-Hispanic/Latino other; percent poverty; percent who work full-time; and percent of women aged 25 and older with less than high school degree, high school degree, some college, or college grad and higher).
ALT TEXT for FIGURE 1, Scatter Chart: The x-axis of the figure displays outcome variables for both matched and non-matched analyses. Outcome variables are presented as follows from left to right across the x-axis: any well-child visit, non-matched analyses; any well-child visit, matched analyses; any ED visit, non-matched analyses; any ED visit, matched analyses; any avoidable ED visit, non-matched analyses; any avoidable ED visit, matched analyses; any ACS hospitalization, non-matched analyses; any ACS hospitalization, matched analyses; follow-up after ED visit, non-matched analyses; follow-up after ED visit, matched analyses; and follow-up after hospitalization, matched analyses. The y-axis conveys odds ratio and 95% confidence intervals; the y-axis number scale starts with zero at the intercept and 3 at the top.

The figure presents odd ratios for the treatment group, defined as children attributed to NCQA recognized medical homes, versus comparison group. The odds ratios are presented as solid squares with vertical lines emanating from the squares that span the 95% confidence interval for each outcome measure. There is also a red horizontal line across the figure that corresponds to an odds ratio of one. If the odds ratio or the 95% confidence interval for any measure touches the horizontal line, then the difference in that outcome measure is not statistically significant between intervention and comparison groups. If the 95% confidence interval is above the red line, the intervention group has a statistically significant higher rate of that outcome compared with the comparison group, and if the 95% confidence interval is below the red line, the intervention group has a statistically significant lower rate of that outcome relative to the comparison group.

In this figure, the 95% confidence intervals for all outcome measures touch the horizontal line, except for 2 measures whose odds ratios and confidence intervals were above the red line, indicating an odds ratio significantly greater than one. These 2 measures are "any ED visit, matched analyses," and "any avoidable ED visit, matched analyses." In addition, the odds ratio and confidence interval for "follow-up after ED visit, matched analyses" was below the red line, indicating an odds ratio significantly less than 1.

Those interested in knowing the exact values of the odds ratios and 95% confidence intervals for all outcome measures may finds these values in Table 3.

Louisiana Results

Descriptive, unadjusted analyses (Table 2) found several expected and unexpected differences between treatment and control groups in Louisiana. For example, relative to the "late recognition" comparison group, treatment group CSHCN had higher rates of any well-child visits (61 percent vs. 48 percent, p=0.02) and treatment group children with hospitalizations had higher rates of follow-up within 30 days (73 percent vs. 61 percent; p=0.03); however, in matched comparison group analyses, a greater percentage of the treatment group had any avoidable ED visit relative to the comparison group (40 percent vs. 35 percent; p<0.01), any ED visit (48 percent vs. 44 percent; p<0.01) and treatment group children with ED visits had lower rates of follow-up within 30 days (42 percent vs. 48 percent; p<0.01).

In multivariable regression results, few outcomes showed significant differences between the treatment group and both comparison groups (Figure 1), and all of them showed the treatment group to have worse outcomes than the comparison groups. In "late recognition" comparison group analyses, there were no treatment-comparison differences in any ED or any avoidable ED visits, but in matched comparison group analyses, treatment children were significantly more likely to have ED and avoidable ED visits (OR=1.19, 95% CI=(1.09, 1.29) and OR=1.26, CI=(1.15, 1.38), respectively). Treatment-comparison differences in outcomes related to care coordination suggested no statistically significant difference in follow-up within 30 days of an ED visit and follow-up within 30 days of hospitalization in "late recognition" analyses. Matched comparison group analyses suggested significantly lower odds of follow-up after ED visit (OR=0.82, 95% CI=(0.74, 0.91)) and no significant difference in follow-up after hospitalization.

New Hampshire Results

Descriptive analyses in Table 2 show that relative to the "late recognition" comparison group, the treatment group was more likely to have a follow-up visit within 30 days of an ED visit (60 percent vs. 49 percent; p=0.01). In matched comparison group analyses, the treatment group was more likely than the comparison group to have a well-child visit (81 percent vs. 76 percent; p<0.01), less likely to experience any avoidable ED visit (28 percent vs. 30 percent; p=0.04) and more likely to receive follow-up after an ED visit (58 percent vs. 52 percent; p<0.01).

The regression analyses showed that treatment group children were significantly more likely than the "late recognition" comparison group to have a follow-up office visit after ED visits (OR=1.56, 95% CI=(1.07, 2.27)). The association between treatment group status and any well-child visits also suggested better outcomes for the treatment group, although the difference was borderline-significant in these analyses (OR=1.35, 95% CI=(1.00, 1.81)). However, the analysis also found the treatment group to be more likely to have an inpatient admission (OR=1.70, 95% CI=(1.06, 2.73)). The regression analysis comparing treatment group children to the matched comparison group supported the favorable findings on any well-child visits (OR=1.29, 95% CI=(1.11, 1.51) and follow-up within 30 days of ED visit (OR=1.30, 95% CI=(1.10, 1.55)) (Figure 2).

FIGURE 2. New Hampshire "Late Recognition" and Matched Comparison Group Analyses: Odds Ratios and 95% Confidence Intervals for Treatment vs. Comparison Group from Logistic Regression Models for Service Utilization and Care Coordinationa
See ALT TEXT at end of this figure.
NOTE:
  1. Models for all outcomes adjusted for age (0-1, 2-5, 6-12 or 13-18 years), sex, number of months enrolled in Medicaid, number of CDPS chronic conditions, number of CDPS prescription drug categories, skeletal and GI CDPS condition categories, CDPS prescription drug categories related to asthma/COPD, ADHD, and inflammatory/autoimmune, combined categories for CDPS condition and prescription drug categories (diagnoses and prescriptions related to cardiovascular disease, diagnoses related to psychiatric conditions or prescriptions for depression/anxiety or psychotic illness/bipolar, diagnoses for CNS conditions or prescriptions for multiple sclerosis/paralysis, and diagnoses related to pulmonary disease or prescriptions for cystic fibrosis, diagnoses or prescriptions related to infectious diseases, including HIV/AIDs or prescriptions for HIV/AIDS, and diagnoses or prescriptions related to renal disease), as well as a series of zip code-level variables from the ACS (percent Hispanic/Latino, non-Hispanic/Latino White, non-Hispanic/Latino Black, non-Hispanic/Latino other; percent poverty; percent who work full-time; and percent of women aged 25 and older with less than high school degree, high school degree, some college, or college grad and higher).
ALT TEXT for FIGURE 2, Scatter Chart: The x-axis of the figure displays outcome variables for both matched and non-matched analyses. Outcome variables are presented as follows from left to right across the x-axis: any well-child visit, non-matched analyses; any well-child visit, matched analyses; any ED visit, non-matched analyses; any ED visit, matched analyses; any avoidable ED visit, non-matched analyses; any avoidable ED visit, matched analyses; any ACS hospitalization, non-matched analyses; any ACS hospitalization, matched analyses; follow-up after ED visit, non-matched analyses; and follow-up after ED visit, matched analyses. The y-axis conveys odds ratio and 95% confidence intervals; the y-axis number scale starts with zero at the intercept and 3.5 at the top.

The figure presents odd ratios for the treatment group, defined as children attributed to NCQA recognized medical homes, versus comparison group. The odds ratios are presented as solid squares with vertical lines emanating from the squares that span the 95% confidence interval for each outcome measure. There is also a red horizontal line across the figure that corresponds to an odds ratio of 1. If the odds ratio or the 95% confidence interval for any measure touches the horizontal line, then the difference in that outcome measure is not statistically significant between intervention and comparison groups. If the 95% confidence interval is above the red line, the intervention group has a statistically significant higher rate of that outcome compared with the comparison group, and if the 95% confidence interval is below the red line, the intervention group has a statistically significant lower rate of that outcome relative to the comparison group.

In this figure, the 95% confidence intervals for all outcome measures touch the horizontal line, except for three outcome measures, whose 95% confidence intervals are above the horizontal red line, indicating odds ratios significantly greater than 1. These 3 measures are "any well-child visit, matched analyses," "any hospitalization, non-matched analyses," and "follow-up after ED visit, non-matched and matched analyses."

Those interested in knowing the exact values of the odds ratios and 95% confidence intervals for all outcome measures may finds these values in Table 4.

Texas Results

Descriptive, unadjusted results in Texas suggested that relative to the "late recognition" comparison group, treatment group children were more likely to have well-child visits (78 percent vs. 59 percent; p<0.01) and more likely to receive follow-up within 30 days of an ED visit (55 percent vs. 38 percent; p<0.01). There were no significant differences for any outcomes between treatment and comparison groups in matched analyses (Table 2).

In multivariable regression analyses, there were few statistically significant treatment-comparison differences in outcome measures (Figure 3). The treatment group had significantly higher odds of any well-child visit (1.73 (1.19, 2.53)) and follow-up after ED visit (2.42 (1.63, 3.61)) when compared to the "late recognition" comparison group, but these differences were not significant in the matched comparison group analyses.

FIGURE 3. Texas "Late Recognition" and Matched Comparison Group Analyses: Odds Ratios and 95% Confidence Intervals for Treatment vs. Comparison Group from Logistic Regression Models for Service Utilization and Care Coordinationa
See ALT TEXT at end of this figure.
NOTE:
  1. Models for all outcomes adjusted for age (0-1, 2-5, 6-12 or 13-18 years), sex, disability status (disabled vs. non-disabled), number of months enrolled in Medicaid, number of CDPS chronic conditions, number of CDPS prescription drug categories, skeletal and GI CDPS condition categories, CDPS prescription drug categories related to asthma/COPD, ADHD, and inflammatory/autoimmune), combined categories for CDPS condition and prescription drug categories (diagnoses and prescriptions related to cardiovascular disease, diagnoses related to psychiatric conditions or prescriptions for depression/anxiety or psychotic illness/bipolar, diagnoses for CNS conditions or prescriptions for multiple sclerosis/paralysis, and diagnoses related to pulmonary disease or prescriptions for cystic fibrosis), as well as a series of zip code-level variables from the ACS (percent Hispanic/Latino, non-Hispanic/Latino White, non-Hispanic/Latino Black, non-Hispanic/Latino other; percent poverty; percent who work full-time; and percent of women aged 25 and older with less than high school degree, high school degree, some college, or college grad and higher).
ALT TEXT for FIGURE 3, Scatter Chart: The x-axis of the figure displays outcome variables for both matched and non-matched analyses. Outcome variables are presented as follows from left to right across the x-axis: any well-child visit, non-matched analyses; any well-child visit, matched analyses; any ED visit, non-matched analyses; any ED visit, matched analyses; any avoidable ED visit, non-matched analyses; any avoidable ED visit, matched analyses; any ACS hospitalization, non-matched analyses; any ACS hospitalization, matched analyses; follow-up after ED visit, non-matched analyses; and follow-up after ED visit, matched analyses. The y-axis conveys odds ratio and 95% confidence intervals; the y-axis number scale starts with zero at the intercept and 4 at the top.

The figure presents odd ratios for the treatment group, defined as children attributed to NCQA recognized medical homes, versus comparison group. The odds ratios are presented as solid squares with vertical lines emanating from the squares that span the 95% confidence interval for each outcome measure. There is also a red horizontal line across the figure that corresponds to an odds ratio of one. If the odds ratio or the 95% confidence interval for any measure touches the horizontal line, then the difference in that outcome measure is not statistically significant between intervention and comparison groups. If the 95% confidence interval is above the red line, the intervention group has a statistically significant higher rate of that outcome compared with the comparison group, and if the 95% confidence interval is below the red line, the intervention group has a statistically significant lower rate of that outcome relative to the comparison group.

In this figure, the 95% confidence intervals for all outcome measures touch the horizontal line, except for 2 measures, whose 95% confidence intervals are above the horizontal red line, indicating odds ratios significantly greater than one. These two measures are "any well-child visit, non-matched analyses" and "follow-up after ED visit, non-matched analyses."

Those interested in knowing the exact values of the odds ratios and 95% confidence intervals for all outcome measures may finds these values in Table 5.

Sensitivity Analyses

None of the sensitivity analyses changed our results in any meaningful way that would suggest either of the comparison approaches we used was biased. In sensitivity analyses in Louisiana, the direction and magnitude of coefficients on treatment status was similar to those from the basic model in both sensitivity analyses on the "late recognition" and matched comparison group analyses (Table 3). Specifically, there were no significant differences between treatment and comparison groups for any outcome measures in "late recognition" analyses, and in both base case and sensitivity analyses using matched samples, treatment children were significantly more likely to have any ED visit, any avoidable ED visit and less likely to have follow-up post-ED visit in almost all models. In matched sub-group analyses that included county as a matching variable, treatment children were significantly less likely to have any well-child visit. Sensitivity analyses in New Hampshire generally result in coefficients with similar magnitude and direction compared to the base case analyses, although there are fewer statistically significant findings in sub-group analyses, likely due to reduced power in the sub-group analyses (Table 4). Sub-group analyses in Texas generally result in similar findings as base case analyses, with the exception of the sensitivity analysis on matched comparison group where we matched on county of residence in 2010 (Table 5). In this sensitivity analysis, treatment group children become significantly less likely to have a well-child visit relative to comparison group children (0.66 (0.44, 0.99)), whereas in the basic model there was no significant treatment-comparison difference in any well-child visit.

TABLE 1. Characteristics of Children with Disabilities and Special Health Care Needs Attributed to NCQA-Recognized and Comparison Providers, by State and Comparison Group
  Non-Matched, "Late Recognition" Comparison Group Analyses Matched Comparison Group Analyses
Louisiana New Hampshire Texas Louisiana New Hampshire Texas
Treatment Group 1a Comparison Group 1b Treatment Group 1a Comparison Group 1b Treatment Group 1a Comparison Group 1b Treatment Group 2c Comparison Group 2d Treatment Group 2c Comparison Group 2d Treatment Group 2c Comparison Group 2d
Providers (N) 114 27 145 54 73 100 111 2,856 141 1,180 57 5,568
Children (N) 9,761 1,343 4,090 777 1,174 1,006 8,412 126,856 2,937 10,574 968 59,622
Weighted Children (N) N/A N/A N/A N/A N/A N/A 8,412 8,336 2,937 2,929 968 961
Demographic Characteristics
Age (%)
   0-1 year 17 11 13 16 15 11 17 18 12 12 16 16
   2-5 years 22 19 18 15 26 17 23 23 19 18 26 27
   6-12 years 37 34 39 36 41 42 37 37 41 42 42 42
   13-18 years 24 35 30 32 18 30 22 22 28 28 16 16
Male (%) 57 54 60 54 59 57 57 57 60 60 59 60
Months Enrolled (mean) 11.4 11.5 11.1 11.0 10.5 10.5 11.5 11.5 11.5 11.5 10.6 10.6
Medicaid Eligibility based on Disability (%) 19 20 <1 <1 25 48 16 16 <1 <1 21 21
Health Status
CDPS Condition Categories (%)
   Cardiovascular 6 6 4 5 8 5 4 4 3 3 6 5
   Psychiatric 33 33 56 50 34 35 33 32 59 60 35 34
   Skeletal 8 9 10 10 9 9 7 7 7 7 7 7
   CNS 6 6 6 4 9 12 4 4 3 3 5 5
   Pulmonary 41 38 27 32 41 33 41 42 26 26 41 41
   GI 13 11 10 12 11 13 12 12 8 8 8 8
   Diabetes 1 2 1 2 1 1 <1 1 <1 <1 <1 <1
   Renal <1 <1 <1 <1 <1 <1 <1 <1 <1 <1 <1 <1
   Substance Abuse <1 1 1 2 <1 <1 <1 <1 <1 <1 <1 <1
   Cancer <1 <1 <1 <1 1 <1 <1 <1 <1 <1 <1 <1
   Developmental Disability 2 2 2 2 3 5 <1 <1 <1 <1 1 1
   Metabolic 2 <1 2 1 2 2 1 1 <1 <1 1 1
   Cerebrovascular <1 <1 <1 <1 1 1 <1 <1 <1 <1 <1 <1
   Hematological 3 2 1 2 3 1 2 2 <1 <1 1 1
   Infectious/ AIDS 6 3 3 4 4 4 6 6 2 2 3 3
Zip-Code Level Socioeconomic Characteristics (mean)
Hispanic/Latino 5 2 4 2 20 30 5 4 4 4 20 46
White, Non-Hispanic/Latino 48 45 90 95 58 49 48 56 90 91 58 41
Black, Non-Hispanic/Latino 42 50 2 1 19 18 42 37 2 1 19 11
All Other, Non-Hispanic/Latino 4 2 4 3 3 3 4 3 4 4 3 3
Poverty: Household income <100% FPL 22 26 10 10 17 19 22 21 10 10 17 23
Women with Less than High School Education 18 20 10 11 17 22 18 19 10 10 17 27
Women with High School Graduate 33 36 31 32 29 29 33 36 32 32 29 29
Women with Some College 22 21 20 20 25 24 22 21 20 20 25 22
Women with College Degree or Higher 27 23 39 38 28 25 27 24 39 38 28 22
Worked full-time, year round 45 41 51 50 47 47 45 45 51 51 47 44
SOURCE: Mathematica Policy Research analyses of 2010 MAX data linked to NCQA data on recognized medical home providers and ACS files for zip code-level characteristics.
NOTES: Demographic and health status characteristics of children were measured in 2010. Zip code-level characteristics were obtained from the ACS 2007-2011 public use files. Available at: http://www.census.gov/acs/www/. Accessed March 28, 2013.
  1. Treatment group 1 comprises all children assigned to providers receiving NCQA-recognition between 2008-2010.
  2. Comparison group 1 comprises children assigned to providers receiving NCQA-recognition between January-October 2011.
  3. Treatment group 2 comprises children assigned to providers receiving NCQA-recognition between 2008-2010 who matched to at least 1 comparison child.
  4. Comparison group 2 comprises children assigned to providers who did not receive NCQA-recognition or did not receive it before November 2011 and who matched treatment group children on all observable characteristics.


TABLE 2. Unadjusted Rates of Service Utilization and Care Coordination of Children Attributed to NCQA-Recognized and Comparison Providers
(%)
  Non-Matched, "Late Recognition" Comparison Group Analyses Matched Comparison Group Analyses
Louisiana New Hampshire Texas Louisiana New Hampshire Texas
Treatment Group 1a Comparison Group 1b Treatment Group 1a Comparison Group 1b Treatment Group 1a Comparison Group 1b Treatment Group 2c Comparison Group 2d Treatment Group 2c Comparison Group 2d Treatment Group 2c Comparison Group 2d
Service Utilization
Any well-child visit 61† 48 80 76 78†† 59 61 63 81†† 76 78 79
Any ED visits 49 48 40 45 43 44 48†† 44 38 40 42 37
Any avoidable ED visits 41 39 30 35 35 34 40†† 35 28† 30 34 29
Any inpatient admission 8 9 5 4 10 9 7 7 3 3 7 6
Any ambulatory care-sensitive admissions 1 1 <1 <1 1 1 1 1 <1 <1 1 1
Care Coordination
Follow-up ambulatory visit within 30 days of ED visit 44 40 60† 49 55†† 38 42†† 48 58†† 52 52 51
Follow-up ambulatory visit within 30 days of inpatient admission 73† 61 87 79 ---e ---e 72 69 84 77 ---e ---e
SOURCE: Mathematica Policy Research analyses of 2010 MAX data linked to NCQA data on recognized medical home providers and ACS files for zip code-level characteristics.
NOTES: Outcomes were measured in 2010.
  1. Treatment group 1 comprises all children assigned to providers receiving NCQA-certification between 2008-2010.
  2. Comparison group 1 comprises children assigned to providers receiving NCQA-certification between January-October 2011.
  3. Treatment group 2 comprises children assigned to providers receiving NCQA-certification between 2008-2010 who matched to at least 1 comparison child.
  4. Comparison group 2 comprises children assigned to providers who did not receive NCQA certification or did not receive it before November 2011 and who matched treatment group children on all observable characteristics.
  5. There were too few children with hospital admissions in New Hampshire and Texas to reliably measure follow-up within 30 days after hospitalization.

† Significantly different from comparison group in unadjusted models that account for clustering of children among providers at the 0.05 level, two-tailed test.
†† Significantly different from comparison group in unadjusted models that account for clustering of children among providers at the 0.01 level, two-tailed test.


TABLE 3. Louisiana Sensitivity Analyses: Odds Ratios and 95% Confidence Intervals for Treatment vs. Comparison Group from Logistic Regression Models for Service Utilization and Care Coordination Compared to Base Case
  Logistic Model 1:
Any Well-Child Visitsa
Logistic Model 2:
Any ED Visitsa
Logistic Model 3:
Any Avoidable ED Visitsa
Logistic Model 4:
Any Inpatient Admissionsa
Logistic Model 5:
Any ACS Admissionsa
Logistic Model 6:
Any Follow-Up Within 30 Days of ED Visita
Logistic Model 7:
Any Follow-Up Within
30 Days of Hospitalizationa
Non-Matched, "Late Recognition" Comparison Group Analyses
Base Caseb 1.50
(0.89, 2.54)
1.01
(0.82, 1.24)
1.05
(0.85, 1.31)
0.76
(0.43, 1.35)
0.87
(0.52, 1.48)
1.05
(0.87, 1.28)
1.62
(0.93, 2.83)
Sensitivity Analysis #1: Limit to children age 2-18 enrolled in Medicaid in 2008. Adjusted for 2008 utilization (any ED visits, hospitalizations, and well-child visits) in all models 1.25
(0.78, 2.02)
0.91
(0.76, 1.09)
0.95
(0.79, 1.15)
0.69
(0.43, 1.12)
0.70
(0.36, 1.34)
0.96
(0.78, 1.18)
1.45
(0.76, 2.73)
Matched Comparison Group Analysis: Odds Ratio for Treatment vs. Comparison
Base Caseb 0.91
(0.70, 1.19)
1.19
(1.09, 1.29)
1.26
(1.15, 1.38)
0.92
(0.80, 1.05)
0.84
(0.67, 1.06)
0.82
(0.74, 0.91)
1.15
(0.91, 1.46)
Sensitivity Analysis #1: Matched on county 0.73
(0.57, 0.93)
1.18
(1.07, 1.30)
1.23
(1.11, 1.36)
1.10
(0.93, 1.29)
1.18
(0.88, 1.57)
0.80
(0.71, 0.90)
1.19
(0.86, 1.75)
Sensitivity Analysis #2: Limited to children ages 2-18, adding any ED visits and well-child visits in 2008 to matching algorithm. Adjusted for 2008 utilization (any ED visits, hospitalizations, and well-child visits) in all models 0.90
(0.69, 1.18)
1.12
(1.02, 1.24)
1.17
(1.06, 1.29)
0.82
(0.67, 0.99)
0.89
(0.66, 1.21)
0.83
(0.73, 0.94)
1.16
(0.78, 1.70)
Sensitivity Analysis #3: Limited to children ages 2-18, adding any ED visits and well-child visits in 2008 to matching algorithm. Also added county of residence in 2010 to matching algorithm. Adjusted for 2008 utilization (any ED visits, hospitalizations and well-child visits) in all models 0.72
(0.56, 0.92)
1.09
(0.96, 1.22)
1.14
(1.01, 1.29)
0.93
(0.72, 1.20)
1.04
(0.66, 1.65)
0.96
(0.83, 1.12)
0.71
(0.39, 1.28)
NOTES:
  1. Models for all outcomes adjusted for age
    (0-1, 2-5, 6-12 or 13-18 years), sex, disability status (disabled vs. non-disabled), number of months enrolled in Medicaid, number of CDPS chronic conditions, number of CDPS prescription drug categories, skeletal and GI CDPS condition categories, CDPS prescription drug categories related to asthma/COPD, ADHD, inflammatory/autoimmune, and nausea), combined categories for CDPS condition and prescription drug categories (diagnoses and prescriptions related to cardiovascular disease, diagnoses related to psychiatric conditions or prescriptions for depression/anxiety or psychotic illness/bipolar, diagnoses for CNS conditions or prescriptions for multiple sclerosis/paralysis, diagnoses related to pulmonary disease or prescriptions for cystic fibrosis, diagnoses related to hematological conditions or prescriptions related to hemophilia/von Willbrands disease, diagnoses or prescriptions related to infectious diseases, including HIV/AIDs or prescriptions for HIV/AIDS, as well as diagnoses or prescriptions related to diabetes, and diagnoses or prescriptions related to renal disease), as well as a series of zip code-level variables from the ACS (percent Hispanic/Latino, non-Hispanic/Latino White, non-Hispanic/Latino Black, non-Hispanic/Latino other; percent poverty; percent who work full-time; and percent of women aged 25 and older with less than high school degree, high school degree, some college, or college grad and higher).
  2. Base case analyses are the same as those presented in Figure 1. They are presented again here to facilitate comparisons with the sensitivity analyses.

Bold font indicates odds ratios that are significantly different from comparison group at the 0.05 level, two-tailed test.


TABLE 4. New Hampshire Sensitivity Analyses: Odds Ratios and 95% Confidence Intervals for Treatment vs. Comparison Group from Logistic Regression Models for Service Utilization and Care Coordination Compared to Base Case
  Logistic Model 1:
Any Well-Child Visitsa
Logistic Model 2:
Any ED Visitsa
Logistic Model 3:
Any Avoidable ED Visitsa
Logistic Model 4:
Any Inpatient Admissionsa
Logistic Model 5:
Any ACS Admissionsa
Logistic Model 6:
Any Follow-Up Within 30 Days of ED Visita
Non-Matched, "Late Recognition" Comparison Group Analyses
Base Caseb 1.35
(1.00, 1.81)
0.89
(0.71, 1.11)
0.84
(0.64, 1.11)
1.70
(1.06, 2.73)
0.79
(0.20, 3.15)
1.56
(1.07, 2.27)
Sensitivity Analysis #1: Limit to children age 2-18 enrolled in Medicaid in 2008. Adjusted for 2008 utilization (any ED visits, hospitalizations, and well-child visits) in all models 1.25
(0.88, 1.77)
0.94
(0,72, 1.22)
0.90
(0.66, 1.21)
1.40
(0.71, 2.76)
--- 1.48
(0.93, 2.36)
Matched Comparison Group Analysis: Odds Ratio for Treatment vs. Comparison
Base Caseb 1.29
(1.11, 1.51)
0.92
(0.84, 1.02)
0.89
(0.79, 1.00)
1.21
(0.91, 1.60)
1.42
(0.72, 2.82)
1.30
(1.10, 1.55)
Sensitivity Analysis #1: Matched on county 1.20
(1.01, 1.42)
0.94
(0.83, 1.06)
0.91
(0.78, 1.05)
0.85
(0.60, 1.22)
1.31
(0.64, 2.71)
1.31
(0.19, 9.02)
Sensitivity Analysis #2: Limited to children ages 2-18, adding any ED visits and well-child visits in 2008 to matching algorithm. Adjusted for 2008 utilization (any ED visits, hospitalizations, and well-child visits) in all models 1.27
(1.08, 1.49)
0.94
(0.84, 1.05)
0.89
(0.79, 1.01)
1.27
(0.93, 1.75)
--- 1.10
(0.93, 1.31)
Sensitivity Analysis #3: Limited to children ages 2-18, adding any ED visits and well-child visits in 2008 to matching algorithm. Also added county of residence in 2010 to matching algorithm. Adjusted for 2008 utilization (any ED visits, hospitalizations and well-child visits) in all models 1.15
(0.96, 1.37)
0.97
(0.85, 1.11)
0.93
(0.80, 1.08)
0.93
(0.61, 1.40)
--- 1.09
(0.90, 1.32)
NOTES:
  1. Models for all outcomes adjusted for age (0-1, 2-5, 6-12 or 13-18 years), sex, number of months enrolled in Medicaid, number of CDPS chronic conditions, number of CDPS prescription drug categories, skeletal and GI CDPS condition categories, CDPS prescription drug categories related to asthma/COPD, ADHD, and inflammatory/autoimmune, combined categories for CDPS condition and prescription drug categories (diagnoses and prescriptions related to cardiovascular disease, diagnoses related to psychiatric conditions or prescriptions for depression/anxiety or psychotic illness/bipolar, diagnoses for CNS conditions or prescriptions for multiple sclerosis/paralysis, and diagnoses related to pulmonary disease or prescriptions for cystic fibrosis, diagnoses or prescriptions related to infectious diseases, including HIV/AIDs or prescriptions for HIV/AIDS, and diagnoses or prescriptions related to renal disease), as well as a series of zip code-level variables from the ACS (percent Hispanic/Latino, non-Hispanic/Latino White, non-Hispanic/Latino Black, non-Hispanic/Latino other; percent poverty; percent who work full-time; and percent of women aged 25 and older with less than high school degree, high school degree, some college, or college grad and higher).
  2. Base case analyses are the same as those presented in Figure 1. They are presented again here to facilitate comparisons with the sensitivity analyses.

Bold font indicates odds ratios that are significantly different from comparison group at the 0.05 level, two-tailed test.


TABLE 5. Texas Sensitivity Analyses: Odds Ratios and 95% Confidence Intervals for Treatment vs. Comparison Group from Logistic Regression Models for Service Utilization and Care Coordination Compared to Base Case
  Logistic Model 1:
Any Well-Child Visitsa
Logistic Model 2:
Any ED Visitsa
Logistic Model 3:
Any Avoidable ED Visitsa
Logistic Model 4:
Any Inpatient Admissionsa
Logistic Model 5:
Any ACS Admissionsa
Logistic Model 6:
Any Follow-Up Within 30 Days of ED Visita
Non-Matched, "Late Recognition" Comparison Group Analyses
Base Caseb 1.73
(1.19, 2.53)
0.87
(0.69, 1.10)
0.85
(0.65, 1.10)
0.88
(0.61, 1.26)
0.85
(0.35, 2.08)
2.42
(1.63, 3.61)
Sensitivity Analysis #1: Limit to children age 2-18 enrolled in Medicaid in 2008. Adjusted for 2008 utilization (any ED visits, hospitalizations and well-child visits) in all models 1.87
(1.22, 2.88)
0.89
(0.69, 1.15)
0.86
(0.62, 1.19)
0.91
(0.59, 1.40)
--- 2.41
(1.52, 3.83)
Matched Comparison Group Analysis: Odds Ratio for Treatment vs. Comparison
Base Caseb 1.04
(0.79, 1.36)
1.00
(0.80, 1.25)
1.04
(0.81, 1.32)
1.31
(0.95, 1.82)
0.86
(0.41, 1.80)
1.26
(0.91, 1.74)
Sensitivity Analysis #1: Matched on county 0.68
(0.46, 0.98)
0.82
(0.64, 1.05)
0.83
(0.63, 1.11)
1.55
(1.00, 2.41)
--- 1.77
(1.13, 2.79)
Sensitivity Analysis #2: Limited to children ages 2-18, adding any ED visits and well-child visits in 2008 to matching algorithm. Adjusted for any ED, any hospitalization, and any well-child visits in 2008 in all models 1.13
(0.83, 1.54)
1.06
(0.83, 1.36)
1.08
(0.83, 1.41)
1.70
(1.09, 2.66)
--- 1.36
(0.99, 1.87)
NOTES:
  1. Models for all outcomes adjusted for age (0-1, 2-5, 6-12 or 13-18 years), sex, disability status (disabled vs. non-disabled), number of months enrolled in Medicaid, number of CDPS chronic conditions, number of CDPS prescription drug categories, skeletal and GI CDPS condition categories, CDPS prescription drug categories related to asthma/COPD, ADHD, and inflammatory/autoimmune), combined categories for CDPS condition and prescription drug categories (diagnoses and prescriptions related to cardiovascular disease, diagnoses related to psychiatric conditions or prescriptions for depression/anxiety or psychotic illness/bipolar, diagnoses for CNS conditions or prescriptions for multiple sclerosis/paralysis, and diagnoses related to pulmonary disease or prescriptions for cystic fibrosis), as well as a series of zip code-level variables from the ACS (percent Hispanic/Latino, non-Hispanic/Latino White, non-Hispanic/Latino Black, non-Hispanic/Latino other; percent poverty; percent who work full-time; and percent of women aged 25 and older with less than high school degree, high school degree, some college, or college grad and higher).
  2. Base case analyses are the same as those presented in Figure 1. They are presented again here to facilitate comparisons with the sensitivity analyses.

Bold font indicates odds ratios that are significantly different from comparison group at the 0.05 level, two-tailed test.


View full report

Preview
Download

"ChildDisV3.pdf" (pdf, 724.89Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®