Age Differences in Racial/Ethnic Disparities in Preventable Hospitalizations for Heart Failure in Connecticut,2009-2015: A Population-Based Longitudinal Study

Abstract

Objective:

Preventable hospitalizations for heart failure result in a large proportion of hospitalizations. The primary objective of this study was to describe longitudinal trends in the association of race/ethnicity with preventable hospitalizations for heart failure in Connecticut and differences in disparities by age.

Methods:

We analyzed data on hospitalizations in all civilian acute-care hospitals in Connecticut during a 7-year period, 2009 through 2015. We used raking methodology to weight the nonhospitalized population to create a reference population representative of the state’s general population. Multivariate regression models examined racial/ethnic disparities among adults aged 35-64, controlling for age, sex, and type of health insurance. For adults aged ≥65, regression models controlled for age and sex.

Results:

After controlling for age and sex, the non-Hispanic black to non-Hispanic white odds ratio for preventable hospitalizations for heart failure ranged from 5.2-6.4 during the study period among adults aged 35-64. Among adults aged ≥65, non-Hispanic black adults had significantly higher odds (range, 1.2-1.8) of preventable hospitalizations than non-Hispanic white adults. Rates among Hispanic adults were significantly higher than rates among non-Hispanic adults after controlling for age and sex among adults aged ≥65 in 2014 and 2015.

Conclusions:

This research provides information for clinical and population-based interventions targeting racial/ethnic gaps in heart failure hospitalizations. Demonstrating the persistent black–white disparity and age differences in racial/ethnic disparities, this study emphasizes the need for focused prevention among vulnerable populations. Raking methodology is an innovative approach to eliminating selection bias in hospital discharge data.

Keywords

racial/ethnic disparities preventable hospitalizations heart failure raking methodology multivariate regression

More than 6 million adults in the United States had heart failure in 2013-2016, an increase from 5.7 million in 2009-2012.¹ Almost half of patients hospitalized for heart failure are likely to die within 5 years after discharge.² Among all cardiovascular conditions, heart failure accounts for the greatest proportion of all-cause hospitalizations in the United States and is a public health concern.³ More than 15 million preventable hospitalizations for heart failure occurred in the United States from 1995 through 2009.⁴

Several studies demonstrated that a patient’s race/ethnicity is a significant predictor of all-cause and preventable hospitalization for heart failure.^5

-9 In 2006, the rate of preventable hospitalizations for heart failure among black persons was 1.62 times that among white persons in Maryland.⁵ Although hospitalization rates for heart failure among Medicare beneficiaries declined for all racial/ethnic groups from 1998 to 2008, rates among African American men decreased at a much lower rate.¹⁰ However, studies on younger populations indicate that crude hospitalization rates for heart failure among younger (aged 18-44) and middle-aged (aged 45-64) black adults are approximately 5-6 times the rates among white adults.⁴

Studies on racial/ethnic disparities typically focus only on single-payer groups, such as Medicare or Medicaid populations, or do not control for payer.^4,5,11 It is important to examine racial/ethnic disparities in multiple-payer groups to account for differences in payer groups. Although state discharge data sets allow for comparisons between multiple health insurance types, most previous studies used only descriptive analyses and included smaller racial/ethnic minority populations (eg, Korean Americans, Native Alaskans).^7
-9 Previous research pointed out the lack of multivariate analysis to examine the role of race/ethnicity in determining preventable hospitalizations. It is also important to examine disparities among non-Hispanic black and Hispanic persons because they account for about 30% of the US population.¹² Furthermore, studies that examined longitudinal trends in racial/ethnic disparities did not consider that the drivers of disparities in heart failure outcomes among adults aged 35-64 and adults aged ≥65 differ; these differences warrant the examination of disparities in preventable hospitalizations separately in these 2 age groups.^4,13

This study fills the gaps in research by using a census of all hospitalizations for heart failure in Connecticut, representing outcomes in a multipayer population. Another important contribution of this study is that it examines racial/ethnic disparities by using a multivariate approach. This approach adds value over past studies that compared crude rates to examine racial/ethnic differences in hospitalization rates. To our knowledge, this study is the first to compare racial/ethnic disparities in preventable hospitalizations for heart failure among adults aged 35-64 and adults aged ≥65 during the past decade. Because preventable hospitalization for heart failure is a population-based metric, regression modeling of discharge data would be inadequate to generate findings that are generalizable.^8,14 Using raking methodology, we assigned weights to the discharge data to generate a reference population representative of the general population, making the results more generalizable.¹⁵

Methods

We obtained data on all hospitalizations occurring in Connecticut from October 1, 2008, through September 30, 2015, from the Connecticut Hospital Inpatient Discharge Database.¹⁶ The Connecticut Department of Public Health maintains the database, which comprises demographic, clinical, and billing data for discharges from all civilian acute-care hospitals in the state. The University of Connecticut Health Center Institutional Review Board and the Connecticut Department of Public Health Human Investigations Committee approved this study.

Measures

Using a case-control design, we examined racial/ethnic disparities in preventable hospitalizations for heart failure in 2 age groups: adults aged 35-64 and adults aged ≥65. The Agency for Healthcare Research and Quality (AHRQ) defines the rate of preventable hospitalization for heart failure as the number of admissions with a principal diagnosis of heart failure per 100 000 population. This definition excludes admissions associated with cardiac procedures, obstetric admissions, and transfers from other institutions.¹⁷ Preventable hospitalization for heart failure, known as the Prevention Quality Indicator 8 (PQI-8), was the primary outcome for this study. The primary predictor of interest for this study was the patient’s race/ethnicity. The race and ethnicity variables in the hospitalization data were self-reported or provider reported. We classified race/ethnicity into 4 categories: non-Hispanic black, non-Hispanic white, Hispanic, and non-Hispanic other race (patients who self-reported as Asian, Pacific Islander, Alaska Native or Native American, and non-Hispanic). Other covariates included health insurance type (Medicare, Medicaid, uninsured, and privately insured) and sex.

Data Screening

We screened discharge data from 1 391 982 discharges among adults aged 35-64 and 1 530 116 discharges among adults aged ≥65 from 29 civilian acute-care hospitals from October 1, 2008, through September 30, 2015, for preventable hospitalizations for heart failure. We split the data set according to patient’s age at discharge (35-64 and ≥65). We excluded data on discharges with missing information on age, transfer to another facility, and source of admission (eg, transfer from another institution). Per AHRQ specifications, we coded as noncases discharges with missing information on patient’s sex, county of residence, quarter of admission, or diagnosis and discharges associated with a major diagnostic code related to pregnancy, childbirth, or puerperium. We also excluded data on discharges with missing information on fiscal year because the primary objective of this study involved analysis of temporal trends.

Weight Adjustment

Our data source consisted of a census of all civilian adults who had PQI-8 hospitalizations in acute-care hospitals in Connecticut. Using US Census data, we then constructed a reference population that was representative of the nonhospitalized population in the state for each year of the study period. Although the marginal totals for the categories of race/ethnicity and health insurance type for the state were available, the joint distributions were not. To construct the reference population, we used raking methodology for weight adjustment. Raking is a simple technique for adjusting the weights of the sample data on the basis of known marginal population characteristics. The weights from the joint race/ethnicity–insurance categories (eg, non-Hispanic black adults using Medicaid) in the sample can be used to approximate the corresponding population weights.¹⁸ We did not use raking on patient sex because the frequency distribution in the study sample was similar to the population of Connecticut. Raking is used extensively in survey research.¹⁵ In 2016, researchers demonstrated this method to be superior to standard weighting in national databases such as the Behavioral Risk Factor Surveillance System.¹⁹

For each age group, we raked 2 categories of data by using the R package anesrake²⁰: race/ethnicity and health insurance type. The raking process converged in <25 iterations for each data set. We assigned a weight of 1 to patients who had any preventable hospitalization events for heart failure during each year. We obtained data on the distribution of race/ethnicity in the population for each year from the bridged estimates of the National Center for Health Statistics published by the Connecticut Department of Health.^21,22 We used data from the American Community Survey on population proportions for each category of health insurance type in each age group.²³ After assigning weights to the reference population to approximate the sample proportions to the race/ethnicity and health insurance type subpopulations in Connecticut, the weighted frequencies were scaled up so that the sum of the weights in the reference population was equal to the total nonhospitalized population of the state for each year and age category.

Statistical Analysis

We used SAS version 9.4 for the descriptive and multivariate logistic regression analysis.²⁴ We calculated sex- and race-stratified PQI-8 rates. We examined temporal trends in crude rates and differences by race/ethnicity in each age group by using linear regression analysis. We estimated multivariate logistic regression models for each age group by using PROC GENMOD with a weight statement to incorporate the assigned weights for race/ethnicity and health insurance type in the analysis, in a stepwise manner for each year. We used the Wald χ² test to determine the significance of the models; a P value <.05 was considered significant.

The equation for the demographic multivariate model for both age groups was

log P /(1 - P {) = β}_{0} {+ β}_{1} (Race/Ethnicity) {+ β}_{2} (Sex) {+ β}_{3} (Age) (centered)

The equation for the socioeconomic model for age 35-64 was

log P /(1 - P {) = β}_{0} {+ β}_{1} (Race/Ethnicity) {+ β}_{2} (Sex) {+ β}_{3} (Payer) {+ β}_{4} (Age) (centered)

where P is the probability of the PQI-8 and β is the coefficient. We constructed standardized deviance residual plots to examine the goodness-of-fit for each model in both age groups.

Results

A large proportion of the adults hospitalized for heart failure in Connecticut were aged ≥65 and non-Hispanic white (Table 1). Among adults aged 35-64, the overall crude annual rate of preventable hospitalizations for heart failure per 100 000 population decreased from 102.8 in 2009 to 83.7 in 2010 and then returned to 101.9 in 2014. Among adults aged ≥65, the overall crude annual rate of preventable hospitalizations for heart failure per 100 000 population declined from 1451.8 in 2009 to 1292.9 in 2012 but approached baseline levels in 2015 (Table 2).

Table 1.

Sociodemographic characteristics of patients hospitalized for heart failure and of reference population, Connecticut, 2009-2015^a

Characteristic	Hospitalized for Heart Failure (PQI-8),^b No. (%)	Reference Population^c
Characteristic	Hospitalized for Heart Failure (PQI-8),^b No. (%)	Unweighted (Nonrepresentative), No. (%)	Weighted (Representative), No. (%)
Aged 35-64
All	9807 (100.0)	622 517 (100.0)	10 040 414 (100.0)
Sex
Female	3712 (37.9)	347 831 (55.9)	5 655 151 (56.3)
Male	6095 (62.1)	274 686 (44.1)	4 385 263 (43.7)
Race/ethnicity
Non-Hispanic black	3004 (30.6)	82 380 (13.2)	931 827 (9.3)
Hispanic	1441 (14.7)	71 345 (11.5)	1 134 723 (11.3)
Non-Hispanic other^d	358 (3.7)	35 584 (5.7)	439 828 (4.4)
Non-Hispanic white	5004 (51.0)	433 208 (69.6)	7 534 037 (75.0)
Aged ≥65
All	45 668 (100.0)	626 863 (100.0)	3 305 646 (100.0)
Sex
Female	25 113 (55.0)	358 230 (57.1)	1 903 085 (57.6)
Male	20 555 (45.0)	268 633 (42.9)	1 402 561 (42.4)
Race/ethnicity
Non-Hispanic black	3504 (7.7)	39 759 (6.3)	205 795 (6.2)
Hispanic	2006 (4.4)	25 961 (4.1)	161 283 (4.8)
Non-Hispanic other^d	1120 (2.5)	22 384 (3.6)	64 332 (1.9)
Non-Hispanic white	39 038 (85.4)	538 759 (86.0)	2 874 237 (87.1)

Abbreviation: PQI-8, Prevention Quality Indicator 8.

^bThe PQI-8 is defined by the Agency for Healthcare Research and Quality as the rate of preventable hospitalizations for heart failure per 100 000 population.¹⁷

^cUsing raking methodology, the nonhospitalized population from the data was assigned weights and scaled up to construct a reference population that was representative of the population of Connecticut in terms of proportion of persons in various demographic categories.

^dPatients who self-reported as Asian, Pacific Islander, Alaska Native or Native American, and non-Hispanic.

Table 2.

Crude annual rates of preventable hospitalizations for heart failure (per 100 000 population), Connecticut, 2009-2015^a

Characteristic	2009		2010		2011		2012		2013		2014		2015
Characteristic	No.	Rate	No.	Rate	No.	Rate	No.	Rate	No.	Rate	No.	Rate	No.	Rate
Aged 35-64
Insurance status
Medicaid	409	488.6	350	358.7	447	414.2	429	360.0	493	416.6	532	337.7	485	298.7
Medicare	454	992.3	372	726.3	477	904.7	435	830.5	468	851.3	469	834.2	491	901.6
Private	515	44.3	406	35.2	404	35.6	408	36.6	443	40.9	417	37.8	426	38.4
Uninsured	80	59.5	80	52.9	57	37.1	68	42.8	60	35.4	56	49.5	45	47.4
Race/ethnicity
Non-Hispanic black	449	364.6	336	256.3	411	309.6	386	286.4	457	335.7	477	347.0	488	349.8
Hispanic	186	136.6	168	110.9	197	125.7	199	121.7	220	129.9	247	140.2	224	122.6
Non-Hispanic other^b	51	91.8	46	78.7	44	73.7	56	89.7	55	84.8	46	66.7	60	85.4
Non-Hispanic white	780	70.2	667	60.0	742	67.3	704	65.0	732	69.3	704	67.3	675	65.6
Sex
Female	546	67.7	439	52.8	496	61.3	509	63.5	565	71.5	597	73.6	560	69.2
Male	920	148.5	778	124.9	898	139.9	836	130.3	899	141.5	877	141.9	887	145.0
Total	1466	102.8	1217	83.7	1394	96.1	1345	93.1	1464	102.7	1474	103.1	1447	101.9
Total age adjusted	—	77.6	—	63.7	—	72.9	—	70.2	—	77.0	—	77.2	—	76.2
Aged ≥65^c
Race/ethnicity
Non-Hispanic black	473	1820.3	375	1318.2	441	1489.7	474	1592.7	516	1664.6	569	1749.6	656	2051.2
Hispanic	267	1361.8	235	1143.8	237	1089.2	276	1151.4	283	1162.0	332	1299.3	376	1367.1
Non-Hispanic other^b	174	2386.0	159	2015.7	139	1627.1	148	1583.8	158	1566.8	158	1542.5	184	1526.2
Non-Hispanic white	5968	1529.0	5315	1286.9	5552	1337.0	5385	1273.4	5490	1290.2	5547	1307.5	5781	1369.7
Sex
Female	3820	1479.4	3348	1232.8	3515	1284.5	3471	1239.8	3541	1255.4	3577	1270.8	3841	1417.5
Male	3062	1655.2	2736	1379.7	2854	1416.3	2812	1365.1	2906	1391.3	3029	1434.9	3156	1486.2
Total	6882	1451.8	6084	1294.8	6369	1340.4	6283	1292.9	6447	1313.2	6606	1341.1	6997	1417.5
Total age adjusted	—	355.3	—	309.3	—	323.4	—	318.7	—	328.0	—	336.8	—	358.1

^bPatients who self-reported as Asian, Pacific Islander, Alaska Native or Native American, and non-Hispanic.

^cAnalysis did not include health insurance type because the vast majority of these adults were covered by Medicare.

Although rates of preventable hospitalizations for heart failure among non-Hispanic white adults were relatively stable over time, rates among non-Hispanic black adults increased from 2010 to 2015 by approximately 36.5% among adults aged 35-64 (from 256.3 to 349.8 per 100 000 population) and by 55.6% among adults aged ≥65 (from 1318.2 to 2051.2 per 100 000 population; Table 2). The gap between Hispanic adults and non-Hispanic white adults among adults aged 35-64 was relatively narrow during the study period, ranging from 50.9 to 72.9 per 100 000 population. Among adults aged ≥65, the difference in rates between Hispanic and non-Hispanic white adults was reversed, ranging from 0.3 to 24.8 per 100 000 population. In the regression analysis, among adults aged 35-64, we found a significant difference between the probability of preventable hospitalizations for heart failure among non-Hispanic black adults and non-Hispanic white adults (P < .001) and between Hispanic adults and non-Hispanic white adults (P = .04). Among adults aged ≥65, the probability of preventable hospitalizations for heart failure differed significantly between non-Hispanic white adults and adults who identified as non-Hispanic other race (P < .001). However, rates by race/ethnicity did not change significantly during the study period for either age group.

In the multivariate logistic regression models examining the association of preventable hospitalizations for heart failure with race/ethnicity while controlling for sex among adults aged 35-64, we found persistent significant differences in the odds between non-Hispanic black adults and non-Hispanic white adults and between Hispanic adults and non-Hispanic white adults (Tables 3 and 4). In the unadjusted model, non-Hispanic black adults had significantly higher odds of preventable hospitalizations for heart failure than did non-Hispanic white adults. Odds ratios (ORs) ranged from 4.3 to 5.3 during the study period: non-Hispanic black adults were as much as 5.3 times more likely than non-Hispanic white adults to have preventable hospitalizations for heart failure. In addition, we found significant differences between non-Hispanic white adults and Hispanic adults: ORs ranged from 1.8 in 2010 to 2.1 in 2014 in the unadjusted model. In the demographic model, the ORs comparing non-Hispanic black adults and non-Hispanic white adults and the ORs comparing Hispanic adults and non-Hispanic white adults were all higher than the ORs comparing these groups in the unadjusted model.

Table 3.

Multivariate logistic regression models examining the role of race/ethnicity on preventable hospitalizations for heart failure among adults aged 35-64, Connecticut, 2009-2012^a

Characteristic	2009	2010	2011	2012
Unadjusted model
Race/ethnicity
Non-Hispanic black	5.2 (4.6-5.9) [<.001]	4.3 (3.8-4.9) [<.001]	4.6 (4.1-5.2) [<.001]	4.4 (3.9-5.0) [<.001]
Hispanic	1.9 (1.7-2.3) [<.001]	1.8 (1.6-2.2) [<.001]	1.9 (1.6-1.9) [<.001]	1.9 (1.6-2.2) [<.001]
Non-Hispanic other^b	1.3 (1.0-1.7) [.06]	1.3 (1.0-1.8) [.07]	1.1 (0.8-1.3) [.56]	1.4 (1.1-1.8) [.02]
Non-Hispanic white	1 [Reference]	1 [Reference]	1 [Reference]	1 [Reference]
Demographic model
Race/ethnicity
Non-Hispanic black	6.2 (5.5-7.0) [<.001]	5.2 (4.6-6.0) [<.001]	5.5 (4.9-6.3) [<.001]	5.4 (4.7-6.1) [<.001]
Hispanic	2.4 (2.1-2.9) [<.001]	2.5 (2.1-2.9) [<.001]	2.5 (2.1-2.9) [<.001]	2.5 (2.1-2.9) [<.001]
Non-Hispanic other^b	1.5 (1.1-2.0) [.01]	1.6 (1.2-2.2) [.002]	1.4 (1.0-1.9) [.045]	1.7 (1.3-2.2) [<.001]
Non-Hispanic white	1 [Reference]	1 [Reference]	1 [Reference]	1 [Reference]
Sex
Female	0.5 (0.5-0.6) [<.001]	0.5 (0.4-0.5) [<.001]	0.5 (0.4-0.6) [<.001]	0.5 (0.5-0.6) [<.001]
Male	1 [Reference]	1 [Reference]	1 [Reference]	1 [Reference]
Age	1.1 (1.1-1.1) [<.001]	1.1 (1.1-1.1) [<.001]	1.1 (1.1-1.1) [<.001]	1.1 (1.1-1.1) [<.001]
Socioeconomic model
Race/ethnicity
Non-Hispanic black	3.6 (3.2-4.0) [<.001]	3.1 (2.7-3.5) [<.001]	3.1 (2.8-3.6) [<.001]	3.1 (2.7-3.5) [<.001]
Hispanic	1.3 (1.1-1.5) [.01]	1.3 (1.1-1.5) [.01]	1.2 (1.1-1.5) [.01]	1.3 (1.1-1.6) [.001]
Non-Hispanic other^b	1.6 (1.2-2.2) [<.001]	1.7 (1.2-2.2) [.001]	1.4 (1.1-2.0) [.02]	1.9 (1.4-2.5) [<.001]
Non-Hispanic white	1 [Reference]	1 [Reference]	1 [Reference]	1 [Reference]
Sex
Female	0.5 (0.5-0.6) [<.001]	0.5 (0.4-0.6) [<.001]	0.5 (0.5-0.6) [<.001]	0.6 (0.5-0.6) [<.001]
Male	1 [Reference]	1 [Reference]	1 [Reference]	1 [Reference]
Age	1.1 (1.1-1.1) [<.001]	1.1 (1.1-1.1) [<.001]	1.1 (1.1-1.1) [<.001]	1.1 (1.1-1.1) [<.001]
Insurance status
Medicaid	10.2 (8.9-11.7) [<.001]	10.2 (8.8-11.9) [<.001]	11.5 (10.0-13.3) [<.001]	9.5 (8.2-10.9) [<.001]
Medicare	16.5 (14.5-18.7) [<.001]	15.4 (13.3-17.8) [<.001]	19.3 (16.8-23.1) [<.001]	17.1 (14.9-19.6) [<.001]
Uninsured	1.2 (0.9-1.5) [.18]	1.4. (1.1-1.8) [.01]	1.0 (0.7-1.3) [.78]	1.1 (0.8-1.4) [.51]
Private	1 [Reference]	1 [Reference]	1 [Reference]	1 [Reference]

^aData source: The Connecticut Hospital Inpatient Discharge Database, which comprises demographic, clinical, and billing data for discharges from all civilian acute-care hospitals in the state.¹⁶ Data included all hospitalizations occurring in Connecticut from October 1, 2008, through September 30, 2012. All values are odds ratio (95% confidence interval) [P value]. The Wald χ² test was used to test models; P < .05 was considered significant.

^bPatients who self-reported as Asian, Pacific Islander, Alaska Native or Native American, and non-Hispanic.

Table 4.

Multivariate logistic regression models examining the role of race/ethnicity on preventable hospitalizations for heart failure among adults aged 35-64, Connecticut, 2013-2015^a

Characteristic	2013	2014	2015
Unadjusted model
Race/ethnicity
Non-Hispanic black	4.9 (4.3-5.5) [<.001]	5.2 (4.6-5.8) [<.001]	5.3 (4.8-6.0) [<.001]
Hispanic	1.9 (1.6-2.2) [<.001]	2.1 (1.8-2.4) [<.001]	1.9 (1.6-2.2) [<.001]
Non-Hispanic other^b	1.2 (0.9-1.6) [.15]	1.0 (0.7-1.3) [.95]	1.3 (1.0-1.7) [.05]
Non-Hispanic white	1 [Reference]	1 [Reference]	1 [Reference]
Demographic model
Race/ethnicity
Non-Hispanic black	5.8 (5.2-6.5) [<.001]	6.1 (5.4-6.9) [<.001]	6.4 (5.7-7.2) [<.001]
Hispanic	2.5 (2.2-2.9) [<.001]	2.7 (2.4-3.2) [<.001]	2.5 (2.1-2.9) [<.001]
Non-Hispanic other^b	1.5 (1.2-2.0) [.002]	1.3 (0.9-1.7) [.12]	1.7 (1.3-2.2) [<.001]
Non-Hispanic white	1 [Reference]	1 [Reference]	1 [Reference]
Sex
Female	0.6 (0.5-0.6) [<.001]	0.6 (0.5-0.7) [<.001]	0.6 (0.5-0.6) [<.001]
Male	1 [Reference]	1 [Reference]	1 [Reference]
Age	1.1 (1.1-1.1) [<.001]	1.1 (1.1-1.1) [<.001]	1.1 (1.1-1.1) [<.001]
Socioeconomic model
Race/ethnicity
Non-Hispanic black	3.4 (3.0-3.9) [<.001]	3.6 (3.2-4.1) [<.001]	3.9 (3.4-4.4) [<.001]
Hispanic	1.4 (1.2-1.6) [<.001]	1.5 (1.3-1.7) [<.001]	1.4 (1.2-1.6) [<.001]
Non-Hispanic other^b	1.6 (1.2-2.1) [<.001]	1.3 (1.0-1.8) [.08]	1.8 (1.4-2.4) [<.001]
Non-Hispanic white	1 [Reference]	1 [Reference]	1 [Reference]
Sex
Female	0.6 (0.5-0.7) [<.001]	0.6 (0.6-0.7) [<.001]	0.6 (0.5-0.7) [<.001]
Male	1 [Reference]	1 [Reference]	1 [Reference]
Age	1.1 (1.1-1.1) [<.001]	1.1 (1.1-1.1) [<.001]	1.1 (1.1-1.1) [<.001]
Insurance status
Medicaid	9.5 (8.3-10.9) [<.001]	8.1 (7.1-9.3) [<.001]	7.0 (6.1-8.0) [<.001]
Medicare	15.4 (13.5-17.6) [<.001]	15.8 (13.8-18.1) [<.001]	16.8 (14.7-19.2) [<.001]
Uninsured	0.8 (0.6-1.1) [.13]	1.2 (0.9-1.6) [.20]	1.1 (0.8-1.5) [.52]
Private	1 [Reference]	1 [Reference]	1 [Reference]

^aData source: The Connecticut Hospital Inpatient Discharge Database, which comprises demographic, clinical, and billing data for discharges from all civilian acute-care hospitals in the state.¹⁶ Data included all hospitalizations occurring in Connecticut from October 1, 2012, through September 30, 2015. All values are odds ratio (95% confidence interval) [P value]. The Wald χ² test was used to test models; P < .05 was considered significant.

^bPatients who self-reported as Asian, Pacific Islander, Alaska Native or Native American, and non-Hispanic.

In the socioeconomic model (Tables 3 and 4), the inclusion of health insurance type in the demographic model substantially affected the ORs. The ORs comparing non-Hispanic black adults with non-Hispanic white adults and the ORs comparing Hispanic adults with non-Hispanic white adults in the socioeconomic model were almost half of those in the demographic model. Medicare and Medicaid enrollees had higher odds of preventable hospitalizations for heart failure than did the privately insured population. The ORs comparing Medicaid adults with privately insured adults fluctuated during the study period, resulting in an overall decrease from 10.2 in 2009 to 7.0 in 2015. Similarly, although the ORs comparing Medicare adults with privately insured adults fluctuated, overall we found a small increase from 16.5 in 2009 to 16.8 in 2015.

For adults aged ≥65, the multivariate logistic regression analysis did not include health insurance type because the vast majority of these adults were covered by Medicare (Tables 5 and 6). Patterns of race/ethnicity in this population were different from patterns among adults aged 35-64. Although in the unadjusted model, non-Hispanic black adults had significantly higher odds of preventable hospitalizations for heart failure than did non-Hispanic white adults for each year of the study period except 2010 (ORs were 1.2 in 2009 and 1.5 in 2015), these ORs were lower than the ORs observed among adults aged 35-64. Except for 2011, we found no significant differences between Hispanic adults and non-Hispanic white adults. In the unadjusted models, adults who self-identified as non-Hispanic other had significantly higher odds of preventable hospitalizations for heart failure than did non-Hispanic white adults in all years except 2015. In the demographic model (adjusting for age and sex), the ORs comparing non-Hispanic black adults with non-Hispanic white adults were generally higher than the ORs in the unadjusted model, and they increased from 1.5 in 2009 to 1.8 in 2015. In 2014 and 2015, Hispanic adults had significantly higher odds than non-Hispanic white adults of preventable hospitalizations for heart failure. We found no unusual patterns in any standardized deviance residual plots.

Table 5.

Multivariate logistic regression models examining the role of race/ethnicity on preventable hospitalizations for heart failure among adults aged ≥65, Connecticut, 2009-2012^a

Characteristic	2009	2010	2011	2012
Unadjusted model
Race/ethnicity
Non-Hispanic black	1.2 (1.1-1.3) [<.001]	1.0 (0.9-1.1) [.65]	1.1 (1.0-1.2) [.03]	1.2 (1.1-1.4) [<.001]
Hispanic	0.9 (0.8-1.0) [.06]	0.9 (0.8-1.0) [.08]	0.8 (0.7-0.9) [.002]	0.9 (0.8-1.0) [.10]
Non-Hispanic other^b	1.6 (1.3-1.8) [<.001]	1.6 (1.4-1.9) [<.001]	1.2 (1.0-1.4) [.02]	1.2 (1.1-1.5) [.01]
Non-Hispanic white	1 [Reference]	1 [Reference]	1 [Reference]	1 [Reference]
Demographic model
Race/ethnicity
Non-Hispanic black	1.5 (1.3-1.6) [<.001]	1.2 (1.1-1.4) [<.001]	1.3 (1.2-1.5) [<.001]	1.5 (1.4-1.7) [<.001]
Hispanic	1.1 (1.0-1.3) [.08]	1.1 (1.0-1.3) [.06]	1.0 (0.9-1.2) [.60]	1.1 (1.0-1.3) [.06]
Non-Hispanic other^b	1.8 (1.5-2.1) [<.001]	1.8 (1.6-2.1) [<.001]	1.4 (1.2-1.7) [<.001]	1.5 (1.2-1.7) [<.001]
Non-Hispanic white	1 [Reference]	1 [Reference]	1 [Reference]	1 [Reference]
Sex
Female	0.8 (0.8-0.8) [<.001]	0.8 (0.8-0.8) [<.001]	0.8 (0.8-0.9) [<.001]	0.8 (0.8-0.9) [<.001]
Male	1 [Reference]	1 [Reference]	1 [Reference]	1 [Reference]
Age	1.1 (1.0-1.1) [<.001]	1.1 (1.1-1.1) [<.001]	1.1 (1.0-1.1) [<.001]	1.1 (1.0-1.1) [<.001]

^aData source: The Connecticut Hospital Inpatient Discharge Database, which comprises demographic, clinical, and billing data for discharges from all civilian acute-care hospitals in the state.¹⁶ Data included all hospitalizations occurring in Connecticut from October 1, 2008, through September 30, 2012. Analysis did not include health insurance type because the vast majority of these adults were covered by Medicare. All values are odds ratio (95% confidence interval) [P value]. The Wald χ² test was used to test models; P < .05 was considered significant.

^bPatients who self-reported as Asian, Pacific Islander, Alaska Native or Native American, and non-Hispanic.

Table 6.

Multivariate logistic regression models examining the role of race/ethnicity on preventable hospitalizations for heart failure among adults aged ≥65, Connecticut, 2013-2015^a

Characteristic	2013	2014	2015
Unadjusted model
Race/ethnicity
Non-Hispanic black	1.3 (1.2-1.4) [<.001]	1.3 (1.2-1.5) [<.001]	1.5 (1.4-1.6) [<.001]
Hispanic	0.9 (0.8-1.0) [.08]	1.0 (0.9-1.1) [.91]	1.0 (0.9-1.1) [.97]
Non-Hispanic other^b	1.2 (1.0-1.4) [.02]	1.2 (1.0-1.4) [.04]	1.1 (1.0-1.3) [.15]
Non-Hispanic white	1 [Reference]	1 [Reference]	1 [Reference]
Demographic model
Race/ethnicity
Non-Hispanic black	1.5 (1.4-1.7) [<.001]	1.6 (1.5-1.7) [<.001]	1.8 (1.6-1.9) [<.001]
Hispanic	1.1 (1.0-1.2) [.10]	1.2 (1.1-1.3) [.01]	1.2 (1.1-1.3) [<.001]
Non-Hispanic other^b	1.4 (1.2-1.6) [<.001]	1.3 (1.1-1.6) [<.001]	1.2 (1.1-1.4) [<.01]
Non-Hispanic white	1 [Reference]	1 [Reference]	1 [Reference]
Sex
Female	0.8 (0.8-0.9) [<.001]	0.8 (0.8-0.8) [<.001]	0.8 (0.8-0.9) [<.001]
Male	1 [Reference]	1 [Reference]	1 [Reference]
Age	1.1 (1.1-1.1) [<.001]	1.0 (1.0-1.1) [<.001]	1.1 (1.0-1.1) [<.001]

^aData source: The Connecticut Hospital Inpatient Discharge Database, which comprises demographic, clinical, and billing data for discharges from all civilian acute-care hospitals in the state.¹⁶ Data included all hospitalizations occurring in Connecticut from October 1, 2012, through September 30, 2015. Analysis did not include health insurance type because the vast majority of these adults were covered by Medicare. All values are odds ratio (95% confidence interval) [P value]. The Wald χ² test was used to test models; P < .05 was considered significant.

^bPatients who self-reported as Asian, Pacific Islander, Alaska Native or Native American, and non-Hispanic.

Discussion

To overcome the lack of generalizability of regression models based exclusively on a sample of hospitalized adults, we adopted a unique approach to weight adjustment for hospital discharge data by creating a reference population of nonhospitalized persons. This reference population was representative of the population of Connecticut in race/ethnicity, health insurance type, and age. The raking methodology used for this weight adjustment is more effective than simple weighting in events where data on the joint distribution of 2 weighted variables are not available.¹⁹

Our analysis showed that non-Hispanic black adults had significantly greater risk than non-Hispanic white adults of preventable hospitalizations for heart failure and that the disparity between these 2 groups persisted over time as rates of preventable hospitalizations for heart failure among non-Hispanic black adults increased from 2010 to 2015. The magnitude of the black–white disparity in our study was similar to the black–white disparity in rates of hospitalizations for heart failure and preventable hospitalizations for heart failure found among adults aged >65 in another study.²⁵ Our study showed interesting differences between adults aged 35-64 and adults aged ≥65 in the role of race/ethnicity in predicting rates of preventable hospitalizations for heart failure. In our study, non-Hispanic black adults had a greater risk than non-Hispanic white adults of preventable hospitalizations for heart failure among those aged <65. It is well-known that the pathology of heart failure and the comorbidity profile differ between young adults and older adults. In addition, the onset of heart failure is earlier among black persons than among white persons,²⁶ and black persons are in general younger than white persons when they are hospitalized for heart failure.²⁷ Cardiovascular disease–specific mortality rates among non-Hispanic black adults aged <65 are up to 2.5 times higher than rates among non-Hispanic white adults.²⁸ The sex-standardized rates of preventable hospitalizations for heart failure in the United States increased from 1995 to 2009 among adults aged 18-44, and a large black–white gap existed.⁴ However, recent research using multivariate approaches to racial/ethnic and payer disparities among younger adults is scarce.

A large proportion of the population aged ≥65 is covered by Medicare, which alleviates some of the problems with health care access experienced by younger adults covered by Medicaid. As age increases, the risk of hospitalization for heart failure increases.²⁹ In addition, comorbidities such as hypertension become more prevalent with age, which further increases the incidence of heart failure.³⁰ The age effect and a more equitable access to health care among adults aged ≥65 may explain the smaller black–white disparity among adults aged ≥65 than among adults aged 35-64 in our study. In addition, our regression modeling of the longitudinal data among adults aged ≥65 showed that ORs for non-Hispanic black adults, compared with non-Hispanic white adults, increased during the 7-year period, thereby widening the black–white disparity. Other studies of Medicare patients showed that although the overall rate of hospitalizations for heart failure decreased among non-Hispanic black adults and non-Hispanic white adults, the rate of decrease was slower among non-Hispanic black adults, thereby widening the black–white gap.¹⁰

Health insurance type played a significant role in rates of preventable hospitalizations for heart failure among adults aged 35-64 in our study. From 1974 to 2004, approximately 80% of heart failure–related hospitalizations occurred among Medicare and/or Medicaid enrollees.³¹ The stepwise approach to our regression analysis showed that differences in insurance type accounted for a large proportion of the race effect among Medicare and Medicaid (ie, dual-eligible) enrollees. The decreasing gap in adjusted odds for preventable hospitalizations for heart failure among Medicaid enrollees and those with private insurance indicates a general downward trend in hospitalizations among Medicaid enrollees after 2011. In 2010, Connecticut was one of the first states to expand Medicaid eligibility to healthy childless adults per the provisions of the Affordable Care Act.³² By 2013, the state’s total Medicaid population increased by 20%; the expanded eligibility attracted healthier and younger persons previously enrolled in private insurance plans and previously eligible unenrolled adults.³³ In our study, Medicare enrollees among adults aged 35-64, on the other hand, were at greater risk than privately insured adults aged 35-64 for preventable hospitalization for heart failure over time. To be eligible for Medicare coverage, adults aged <65 must be disabled for at least 24 months or must have a debilitating condition (such as end-stage renal disease); both comorbidities increase the risk of hospitalization.³⁴ Poorer medication adherence behaviors among Medicaid enrollees, compared with medication adherence behaviors among those with private insurance, may explain the higher risk of preventable hospitalization for heart failure.³⁵ The Medicaid population also has limited access to outpatient care; this limited access could affect outcomes related to ambulatory care–sensitive conditions such as heart failure.³⁶

Limitations

This study had several limitations. First, we lacked data on hospitalizations in Veterans Administration facilities and on hospitalizations of Connecticut residents in facilities outside the state. Hence, we limited our analysis to the civilian population. Connecticut, like all northeastern states, experiences the “snowbird” phenomenon, whereby many adults aged ≥55 migrate to southern states, such as Florida, for >30 days per year.³⁷ Our data set did not capture hospitalizations among these adults aged ≥55 while they were out of state. Second, data on race/ethnicity were either self-reported or provider-reported and may not be completely accurate. However, in previous studies, we demonstrated >96% agreement of reported race across multiple admissions and consistency with the state’s racial/ethnic population distribution.³⁸ Racial/ethnic measures need to be enhanced to further understand hospitalizations among multiracial persons. Third, we could not rake the actual age of patients because we lacked population data in age categories shorter than 5-year intervals. Finally, because of limitations of the data, we could not examine the role of dual-eligible status, socioeconomic factors, disease severity, or comorbidities in influencing racial/ethnic disparities in rates of preventable hospitalizations for heart failure.

Conclusions

Methodologically, our study presented an innovative approach to eliminating selection bias in hospitalization data. We constructed a reference population representative of a state by using a raking methodology when data on joint distribution of race/ethnicity and insurance type were not available. This novel approach could be applied to state and national data sets, including state hospital discharge data sets, the National Inpatient Sample, and National Hospital Discharge Survey, to construct population-level models of risk factors.

Identifying risk factors among age groups is important for efficient clinical care and public health interventions. Overall, we found little change in the rates of preventable hospitalizations for heart failure among adults aged 35-64 and adults aged ≥65 from 2009 to 2015. However, we found a persistent and widening black–white gap. The race/ethnicity effect was more prominent among adults aged 35-64 than among adults aged ≥65. Our study contributes to public health by examining differences in racial/ethnic disparities in rates of preventable hospitalizations for heart failure by age. This research can inform population-based interventions targeting racial/ethnic gaps in preventable hospitalizations for heart failure.

Footnotes

Acknowledgments

The authors thank the Connecticut Department of Public Health for providing access to the Connecticut Hospital Inpatient Discharge Database. The authors assume full responsibility for all analyses, interpretations, and conclusions. The Connecticut Department of Public Health does not endorse or assume any responsibility for any analyses, interpretations, or conclusions based on the data. The authors are grateful to colleagues in the Department of Statistics for their help in data cleaning and Gregory Matthews, PhD, for his input on methods.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Riddhi P. Doshi, PhD, MPH, MBBS

References

Benjamin

Muntner

Alonso

, et al. Heart disease and stroke statistics—2019 update: a report from the American Heart Association. Circulation. 2019;139(10):e56–e528. doi:10.1161/CIR.0000000000000659

Mozaffarian

Roger

, et al. Heart disease and stroke statistics—2013 update: a report from the American Heart Association. Circulation. 2013;127(1):e6–e245. doi:10.1161/CIR.0b013e31828124ad

Fida

Piña

. Trends in heart failure hospitalizations. Curr Heart Fail Rep. 2012;9(4):346–353. doi:10.1007/s11897-012-0117-5

Will

Valderrama

Yoon

. Preventable hospitalizations for congestive heart failure: establishing a baseline to monitor trends and disparities. Prev Chronic Dis. 2012;9:E85. doi:10.8888/pcd9.110260

O’Neil

Lake

Merrill

Wilson

Mann

Bartnyska

. Racial disparities in hospitalizations for ambulatory care–sensitive conditions. Am J Prev Med. 2010;38(4):381–388. doi:10.1016/j.amepre.2009.12.026

Bindman

Grumbach

Osmond

, et al. Preventable hospitalizations and access to health care. JAMA. 1995;274(4):305–311. doi:10.1001/jama.1995.03530040033037

Gounder

Seeman

Holman

, et al. Potentially preventable hospitalizations for acute and chronic conditions in Alaska, 2010-2012. Prev Med Rep. 2016;4:614–621. doi:10.1016/j.pmedr.2016.03.017

Heo

H-H

Sentell

Ahn

Miyamura

Braun

. Disparities in potentially preventable hospitalizations for chronic conditions among Korean Americans, Hawaii, 2010-2012. Prev Chronic Dis. 2015;12:E152. doi:10.5888/pcd12.150057

Sentell

Miyamura

Ahn

Chen

Seto

Juarez

. Potentially preventable hospitalizations for congestive heart failure among Asian Americans and Pacific Islanders in Hawai’i. J Immigr Minor Health. 2015;17(5):1289–1297. doi:10.1007/s10903-014-0098-4

10.

Chen

Normand

S-LT

Wang

Krumholz

. National and regional trends in heart failure hospitalization and mortality rates for Medicare beneficiaries, 1998-2008. JAMA. 2011;306(15):1669–1678. doi:10.1001/jama.2011.1474

11.

Braunstein

Anderson

Gerstenblith

, et al. Noncardiac comorbidity increases preventable hospitalizations and mortality among Medicare beneficiaries with chronic heart failure. J Am Coll Cardiol. 2003;42(7):1226–1233. doi:10.1016/s0735-1097(03)00947-1

12.

Will

Nwaise

Schieb

Zhong

. Geographic and racial patterns of preventable hospitalizations for hypertension: Medicare beneficiaries, 2004-2009. Public Health Rep. 2014;129(1):8–18. doi:10.1177/003335491412900104

13.

Wong

Hawkins

Petrie

, et al. Heart failure in younger patients: the Meta-Analysis Global Group in Chronic Heart Failure (MAGGIC). Eur Heart J. 2014;35(39):2714–2721. doi:10.1093/eurheartj/ehu216

14.

Freedland

Rich

Skala

Carney

Dávila-Román

Jaffe

. Prevalence of depression in hospitalized patients with congestive heart failure. Psychosom Med. 2003;65(1):119–128. doi:10.1097/01.PSY.0000038938.67401.85

15.

Dal Grande

Chittleborough

Campostrini

Tucker

Taylor

. Health estimates using survey raked-weighting techniques in an Australian population health surveillance system. Am J Epidemiol. 2015;182(6):544–556. doi:10.1093/aje/kwv080

16.

Connecticut Department of Public Health, Office of Health Care Access, Data Sources and Limitations. Statewide Healthcare Facilities and Services Plan. Hartford, CT: Connecticut Department of Public Health; October 2012. https://portal.ct.gov/-/media/OHS/ohca/HC_Facilities_Advisory_Body/OHCAStatewideFacilitiesandServicesChapter11DataSourcesandLimitationspdf.pdf ?la=en. Accessed September 9, 2019.

17.

Agency for Healthcare Research and Quality. AHRQ quality indicators (AHRQ QI) ICD-9-CM and ICD-10-CM/PCS specification enhanced version 5.0. https://www.qualityindicators.ahrq.gov/Downloads/Modules/PQI/V50-ICD10/TechSpecs/PQI%2008%20Heart%20Failure%20Admission%20Rate.pdf. Published 2015. Accessed December 4, 2018.

18.

Lumley

. Complex Surveys: A Guide to Analysis Using R. Vol 565. Hoboken, NJ: John Wiley & Sons; 2011.

19.

Iachan

Pierannunzi

Healey

Greenlund

Town

. National weighting of data from the Behavioral Risk Factor Surveillance System (BRFSS). BMC Med Res Methodol. 2016;16(1):155. doi:10.1186/s12874-016-0255-7

20.

Pasek

. Package “anesrake.” ANES raking implementation, version 0.80. https://cran.r-project.org/web/packages/anesrake/anesrake.pdf. Published 2018. Accessed December 4, 2018.

21.

Ingram

Parker

Schenker

, et al. United States Census 2000 population with bridged race categories. Vital Health Stat 2. 2003;(135):1–55.

22.

Connecticut Department of Public Health. Annual state & county population with demographics. http://www.ct.gov/dph/cwp/view.asp ?a=3132&q=388 152. Published 2017. Accessed October 5, 2017.

23.

American Community Survey. Health insurance coverage status by sex by age: 2009-2015 American Community Survey 1-year estimates. https://factfinder.census.gov/faces/tableservices/jsf/pages/productview.xhtml ?pid=ACS_16_1YR_B27001&prodType=table . Accessed November 24, 2018.

24.

SAS [computer program]. Version 9.4. Cary, NC: SAS Institute Inc; 2014.

25.

Alexander

Grumbach

Remy

Rowell

Massie

. Congestive heart failure hospitalizations and survival in California: patterns according to race/ethnicity. Am Heart J. 1999;137(5):919–927. doi:10.1016/s0002-8703(99)70417-5

26.

Bibbins-Domingo

Pletcher

Lin

, et al. Racial differences in incident heart failure among young adults. N Engl J Med. 2009;360(12):1179–1190. doi:10.1056/NEJMoa0807265

27.

Afzal

Ananthasubramaniam

Sharma

, et al. Racial differences in patients with heart failure. Clin Cardiol. 1999;22(12):791–794. doi:10.1002/clc.4960221207

28.

Jolly

Vittinghoff

Chattopadhyay

Bibbins-Domingo

. Higher cardiovascular disease prevalence and mortality among younger blacks compared to whites. Am J Med. 2010;123(9):811–818. doi:10.1016/j.amjmed.2010.04.020

29.

Haldeman

Croft

Giles

Rashidee

. Hospitalization of patients with heart failure: national hospital discharge survey, 1985 to 1995. Am Heart J. 1999;137(2):352–360. doi:10.1053/hj.1999.v137.95495

30.

Levy

Larson

Vasan

Kannel

. The progression from hypertension to congestive heart failure. JAMA. 1996;275(20):1557–1562.

31.

Fang

Mensah

Croft

Keenan

. Heart failure-related hospitalization in the U.S., 1979 to 2004. J Am Coll Cardiol. 2008;52(6):428–434. doi:10.1016/j.jacc.2008.03.061

32.

Louise

. Connecticut and the ACA’s Medicaid expansion. Healthinsurance.org: a trusted independent health insurance guide since 1994. https://www.healthinsurance.org/connecticut-medicaid. Accessed September 10, 2019.

33.

Sommers

Kenney

Epstein

. New evidence on the Affordable Care Act: coverage impacts of early Medicaid expansions. Health Aff (Millwood). 2014;33(1):78–87. doi:10.1377/hlthaff.2013.1087

34.

US Department of Health and Human Services. Who is eligible for Medicare? https://www.hhs.gov/answers/medicare-and-medicaid/who-is-elibible-for-medicare/index.html. Published 2014. Accessed November 4, 2017.

35.

Esposito

Bagchi

Verdier

Bencio

Kim

. Medicaid beneficiaries with congestive heart failure: association of medication adherence with healthcare use and costs. Am J Manag Care. 2009;15(7):437–445.

36.

Medicaid Access Study Group. Access of Medicaid recipients to outpatient care. N Engl J Med. 1994;330(20):1426–1430. doi:10.1056/NEJM199405193302007

37.

Smith

House

. Snowbirds, sunbirds, and stayers: seasonal migration of elderly adults in Florida. J Gerontol B Psychol Sci Soc Sci. 2006;61(5):S232–S239. doi:10.1093/geronb/61.5.s232

38.

Aseltine

Jr Yan

Fleischman

Katz

DeFrancesco

. Racial and ethnic disparities in hospital readmissions after delivery. Obstet Gynecol. 2015;126(5):1040–1047. doi:10.1097/AOG.0000000000001090