Spatial Associations of Multiple Chronic Conditions Among Older Adults

Abstract

Multimorbidity, the presence of two or more chronic conditions in an individual, presents a major challenge for meeting the health care needs of older adults. This study advances understanding of multiple chronic conditions by using local colocation quotients to reveal spatial associations for five chronic conditions (arthritis, diabetes, heart disease, hypertension, and pulmonary disease) in a statewide panel of older adults in New Jersey. Among adults with three or more conditions, large concentrations of Arthritis-Heart Disease-Pulmonary Disease, Arthritis-Hypertension-Pulmonary Disease, and Diabetes-Heart Disease-Hypertension were observed, each triad located in different regions of the state. Individuals with other triads of conditions, in contrast, were distributed among all older adults in the sample as expected with no areas of local concentration. The study provides gerontologists with a new and effective method for uncovering geographical patterns in combinations of chronic conditions among the populations they serve, thereby enabling more effective interventions.

Keywords

older adults chronic disease multimorbidity spatial association joint disease mapping

Advances in medical knowledge and treatment over the last century contributed to the development of clinical and public health programs focused on single diseases. The same advances in diagnosis and care, coupled with better understanding of the broader contexts affecting individuals’ health and access to databases of individual health status, also drew attention to people suffering from multiple health problems. As early as the 1970s, practitioners recognized the need to develop better methods for representing the interrelationships and effects of multiple diseases to accurately assess their occurrence and evaluate treatments (Feinstein, 1970).

As initially framed, the challenge in addressing multiple diseases involved accounting for comorbid conditions in addition to a single health problem of interest, referred to as the “index” disease or presenting condition affecting a patient. More recently, practitioners have made a distinction between comorbidity and multimorbidity, defined as the presence of two or more chronic conditions in an individual (van den Akker, Buntinx, & Knottnerus, 1996). Studies of multimorbidity show that both children and adults are affected (Rocca et al., 2014). Because the elderly population is increasing in all regions of the world (United Nations Department of Economic and Social Affairs, Population Division, 2013) and the prevalence of multiple chronic conditions increases with age (Barnet et al., 2012; Ornstein, Nietert, Jenkins, & Litvin, 2013; van den Akker, Buntinx, Metsemakers, Roos, & Knottnerus, 1998), a growing body of research highlights multimorbidity in older adults. The occurrence of multiple chronic conditions in this age group is clearly a global health problem leading to complications in treatment approaches and high cost burdens on health care systems (Caughey, Vitry, Gilbert, & Roughead, 2008; Fortin, Bravo, Hudon, Vanasse, & Lapointe, 2005; Kowal, Williams, Wu Fan, Arokiasamy, & Chatterji, 2012; Lima et al., 2009; Marengoni et al., 2011; Oni et al., 2014; Wolff, Starfield, & Anderson, 2002). Two thirds of health care costs in the U.S. population are tied to the treatment of people with multiple morbidities (LeRoy et al., 2014).

This study advances our understanding of multimorbidity in older adults by using local colocation quotients to reveal spatial associations among chronic conditions in a statewide panel of individuals aged 50 to 74 in New Jersey. Most maps of the distribution of people with multiple health conditions display the prevalence of individuals with high numbers of chronic conditions, regardless of the specific conditions (Lochner & Shoff, 2015). Yet, recent work indicates that specific combinations of chronic conditions carry unique implications for individual-level outcomes (Pruchno, Wilson-Genderson, & Heid, 2016). There is a pressing need to understand the specific combinations of chronic conditions observed in particular populations and compare them with the multimorbidities observed in other localities, both to inform medical and public health practice and to investigate the etiology of multiple chronic conditions (Rocca et al., 2014).

Early maps of disease focused primarily on infectious diseases such as yellow fever and cholera and revealed spatial patterns in conditions at times when the etiologies of these diseases were poorly understood (Gilbert, 1958; Shannon, 1981). When heart disease, stroke, and other chronic diseases emerged as leading causes of mortality and morbidity in industrialized countries, mapping and spatial analysis were adopted as methods for documenting geographical patterns of chronic disease (Howe, 1963; National Cancer Institute Epidemiology Branch, 1975). Efforts to identify areas where particular chronic diseases are prevalent continue to this day (Noble, Smith, Mathus, Robson, & Greenhalgh, 2012). The Division for Heart Disease and Stroke Prevention in the U.S. Centers for Disease Control and Prevention funds collaborative training projects with states to use geographic information systems (GIS) surveillance and mapping to document the burden of chronic diseases, inform program and policy development, and enhance partnerships to reduce health disparities in chronic conditions (National Center for Chronic Disease Prevention and Health Promotion, 2016).

Yet, recent applications of mapping and spatial analysis in studies of chronic disease continue to focus primarily on single conditions. Although recommendations for research to advance our understanding of multiple chronic conditions stress the importance of “considering the person in the context of their relationships and community” (LeRoy et al., 2014), little research has addressed the complex geography of multiple chronic conditions.

In this study, spatial associations among five chronic conditions—arthritis, diabetes, heart disease, hypertension, and pulmonary disease—in older adults diagnosed with at least three of these conditions are calculated and mapped to identify any local areas where the prevalence of specific triads of conditions is greater than in the state as a whole. Research into joint mapping of multiple chronic conditions within populations of older adults complements other studies reporting the frequency and prevalence of different combinations of chronic health problems for populations as a whole (Meraya, Raval, & Sambamoorthi, 2015; Rocca et al., 2014) and highlights potential disparities or unique geographic need for intervention.

Older adults with multiple chronic conditions have greater self-care needs, and the burden of managing multimorbidity by individuals and their caregivers raises important questions about the design of person- and caregiver-centered service (Boyd & Fortin, 2010; May et al., 2014). For some chronic conditions, the goals of care are shared such that treatment recommended for one condition results in better care for the older adult’s other chronic conditions. Treatment goals for hypertension and certain kinds of heart disease, for example, are shared with those for diabetes (Magnan et al., 2015). Other combinations may include conditions that do not share treatment goals. The management of care is more challenging in the latter situation because there may be a greater number of different treatments required, and some of these may conflict. Identifying local areas where older adults have been diagnosed with particular combinations of chronic conditions and face an expanding array of health management tasks can contribute to gerontologic practice at several levels. Information about the specific associations of conditions within localities highlights the needed self-care functions in the community so that practitioners can provide support to older adults and their caregivers. At the same time, practitioners can work to improve the design of service systems so that they improve the capacity of older adults and their caregivers to obtain the care they need to age successfully (May et al., 2014).

Method

Data

The Ongoing Research on Aging in New Jersey–Bettering Opportunities for Wellness in Life (ORANJ BOWL) panel provided data for the study. In 2011, New Jersey and Florida had the highest prevalence of beneficiaries with two or more chronic conditions (75%) and with six or more chronic conditions (18%), based on beneficiary data from the Medicare program (Lochner, Goodman, Posner, & Parekh, 2013).

The ORANJ BOWL panel includes 5,688 older people, interviewed between November 2006 and April 2008. Inclusion criteria required that participants be between the ages of 50 and 74, living in New Jersey, and able to participate in a 1-hour English language telephone interview. Panel members were recruited by telephone cold calling using list-assisted random-digit-dialing (LA-RDD) procedures. Coverage of residential plain old telephone service (POTS) numbers for the population represented by the panel’s sample is estimated as 95%. The demographics of this sample make coverage loss due to the growing number of cell phone-only households very small (Blumberg & Luke, 2007). Based on American Association for Public Opinion Research standards, ORANJ BOWL achieved a response rate of 58.76%, and a cooperation rate of 72.9%, rates consistent with or higher than the average response rates in RDD efforts during this same time.

Comparison of characteristics of ORANJ BOWL respondents with those of all persons age 50 to 74 living in New Jersey reveals that they have similar racial composition, rates of being born in the state, and marital status. The ORANJ BOWL sample has a higher proportion of women (63.7% to 53.3%) and a higher percentage of individuals with advanced secondary degrees (18.5% to 14.8%). It underrepresents persons of Hispanic descent, as participants were restricted to those fluent in English (Pruchno, Wilson-Genderson, Rose, & Cartwright, 2010).

The ORANJ BOWL survey is representative of the geographical distribution of population in the state, with large numbers in the northeast around Newark in the New York metropolitan area and in the east central region near Philadelphia. The absence of ORANJ BOWL participants in the south central area of the state reflects the lack of population in the New Jersey Pine Barrens ecological zone.

The geographic unit of analysis for this study is the residential location of the older adult. To protect confidentiality of the data for the purposes of this analysis, Census block of residence, based on 2000 Census data, was determined from each participant’s mailing address to aid in geospatial analysis and mapping. Because they were missing Census block data, 112 ORANJ BOWL participants were excluded from the sample for this study. Data on the chronic conditions of interest were complete for the remaining 5,576 participants. Approval for human participant research was granted by the University of Medicine and Dentistry of New Jersey. The protocol has been reassessed and approved annually by the institutional review board (IRB) of Rowan University due to university reorganization.

Measures

Residential location

The centroid of the Census block of residences was used to represent the residential location of the older adult, a commonly used form of data masking to protect confidentiality (Rushton et al., 2008). The Census block level is the basic reporting level for population data, and Census blocks are the smallest area units in the Census hierarchy, nested in block groups and tracts. Because the ORANJ BOWL surveys were conducted from 2006 to 2008, Census blocks defined for the 2000 U.S. Census were used. Census geographic data are maintained and published using geographic coordinates (lon, lat). These data were projected to the New Jersey State Plane Coordinate System using ArcGIS 10.2, and the residential location of each participant was approximated with the state plane coordinates of the block centroid of residence.

Chronic conditions

The ORANJ BOWL survey asked participants to self-report whether or not a physician had ever told them that they have any of 14 chronic conditions. Questions of this type are used in the National Health Information Survey (NHIS) and other similar surveys (National Center for Health Statistics, 2016). Five chronic conditions were selected for analysis: arthritis, diabetes, heart disease, hypertension, and pulmonary disease. These five conditions are among the 20 selected by the U.S. Department of Health and Human Services Office of the Assistant Secretary of Health as part of a strategic framework for improving the nation’s response to the challenge of multiple chronic conditions (Goodman, Posner, Huang, Parekh, & Koh, 2013), and arthritis, diabetes, heart disease, and hypertension are among the most prevalent diseases among U.S. adults (Ward, Schiller, & Goodman, 2014). Arthritis, heart disease, and diabetes are leading causes of death and disability (Ford, Croft, Posner, Goodman, & Giles, 2013), with combinations of heart disease and hypertension and diabetes and hypertension among the most common multimorbidities (Freid, Bernstein, & Bush, 2012).

This analysis was designed to detect spatial associations in different triads of these five chronic conditions among older adults with three, four, or five conditions. Of the 5,576 participants with residential locations, 993 (18%) reported having three or more of the five conditions (Table 1), but their 3,335 conditions accounted for almost half (44%) of the total number of the selected chronic conditions reported by the 5,576 participants. This association between the number of people who have multiple chronic conditions and their share of the burden of chronic disease in the population as a whole is commonly observed in studies of multimorbidities (Centers for Medicare and Medicaid Services, 2012; Lehnert et al., 2011).

Table 1.

Multiple Chronic Condition Prevalence Among Older Adults in New Jersey Panel.

	Adults (n = 5,576)
	N
None
No chronic condition	1,530
One
Arthritis	622
Diabetes	86
Heart Disease	121
Hypertension	766
Pulmonary Disease	210
Total	1,805
Two
Arthritis-Diabetes	53
Arthritis-Heart Disease	74
Arthritis-Hypertension	512
Arthritis-Pulmonary Disease	132
Diabetes-Heart Disease	24
Diabetes-Hypertension	171
Diabetes-Pulmonary Disease	8
Heart Disease-Hypertension	130
Heart Disease-Pulmonary Disease	18
Hypertension-Pulmonary Disease	125
Total	1,247
Three
Arthritis-Diabetes-Heart Disease	18
Arthritis-Diabetes-Hypertension	163
Arthritis-Diabetes-Pulmonary Disease	25
Arthritis-Heart Disease-Hypertension	135
Arthritis-Heart Disease-Pulmonary Disease	39
Arthritis-Hypertension-Pulmonary Disease	186
Diabetes-Heart Disease-Hypertension	55
Diabetes-Heart Disease-Pulmonary Disease	8
Diabetes-Hypertension-Pulmonary Disease	30
Heart Disease-Hypertension-Pulmonary Disease	36
Total	695
Four
Arthritis-Diabetes-Heart Disease-Hypertension	77
Arthritis-Diabetes-Heart Disease-Pulmonary Disease	9
Arthritis-Diabetes-Hypertension-Pulmonary Disease	62
Arthritis-Heart Disease-Hypertension-Pulmonary Disease	66
Diabetes-Heart Disease-Hypertension-Pulmonary Disease	26
Total	240
Five
Arthritis-Diabetes-Heart Disease-Hypertension-Pulmonary Disease	58

Analysis Plan

Several methods for analyzing spatial association are available. In their work introducing the Colocation Quotient (CLQ), Leslie and Kronenfeld (2011) review the join count statistic and the cross-k-function. The CLQ provides a summary measure of whether observations in pairs of categories are spatially associated or not in the entire study area. The CLQ, unlike other measures of spatial association, takes the geometric pattern of the joint categories as a given.

Subsequent research generalized the CLQ and introduced a local measure of colocation (R. G. Cromley, Hanink, & Bentley, 2014). Local statistics summarize data for individual observations within a larger study area (Lloyd, 2011). A modified version of the Local Colocation Quotient (LCLQ) introduced by R. G. Cromley et al. (2014) was developed to calculate the measures of association of specific triads of chronic conditions among older adults with three or more conditions.

The LCLQ provides a measure of whether or not observations in pairs of categories are associated in the vicinity or neighborhood of every observation. In this study, the categories were the five chronic conditions of interest and the presence or absence of each condition in every older adult in the sample was known. LCLQs have been used to study patterns of successful aging (E. K. Cromley, Wilson-Genderson, Christman, & Pruchno, 2015).

The application of the LCLQ is straightforward when every individual has only a single condition or not. If every older adult has no or only one chronic condition, the LCLQ can be calculated by finding the number of older adults within a particular neighborhood of each older adult in the sample and finding the ratio of observed to expected adults with the same chronic disease or another type of chronic disease within that neighborhood. The expected number of adults with a particular chronic condition is based on the frequency of the condition in the entire study population. The neighborhood may be defined in one of two ways, using either a fixed filter based on distance from each older adult or a spatially adaptive filter based on k-order nearest neighbors.

In this study, the LCLQ methodology must deal with the distribution of multiple conditions across the distribution of older adults who have different numbers of chronic conditions. Older adults with more than one condition fall into more than one category. Therefore, two steps were needed. In the first step, the k number of nearest older adult neighbors for each older person was identified. In the second step, the presence of each chronic condition reported for each of those older adults—referred to here as a “person-condition”—among the nearest neighbors was assessed. Person-conditions are distributed across the underlying distribution of older adults with multiple chronic conditions. The expected number of adults with a particular chronic condition was based on the frequency of the condition in the entire study population.

The algorithm for calculating LCLQs for each of the 993 older adults with three to five chronic conditions required two data inputs. The first database was a table of 993 records, one for each older adult, with an integer index from 1 to 993, a unique numeric ID, and the x and y coordinates of the block centroid of residence. In this database, a person was included only once. The second database was a table of the 3,335 “person-conditions,” one record for each chronic condition of interest of the 993 older adults. In this database, a person is included three, four, or five times depending on the presence of each of the five chronic conditions. Each person-condition record has the integer index for the older adult, an integer numeric code for one of the five chronic conditions the older adult has, and the x and y coordinates of the block centroid of residence.

A Statistical Analysis System (SAS) program, Version 9.3 using proc iml, was written to calculate the LCLQs for each of the 3,335 person-conditions with each of the five chronic conditions of interest. A copy of this program with information on preparing the input data sets is available from the corresponding author. In the first step of the program, the 100 nearest neighbors of each older adult were identified by calculating the Euclidean distance from every older adult to every other older adult based on residential block coordinates in the first database table. This distance set the variable bandwidth distances for calculating the LCLQs of the person-conditions. Once the bandwidths were calculated, the LCLQ was calculated for the table of person-conditions using Gaussian spatial weights according to the following formula:

{L C L Q}_{i} = P N (b | a_{i}) / P (b | a) .

The numerator PN(b|a_i) is the probability, given the location of a particular a-point, an individual with chronic disease a, of finding a b-point for an individual who has chronic disease b within the particular a-point’s neighborhood:

P N (b | a_{i}) = \sum_{j ϵ D j \neq i} w_{i j} x_{j} / \sum_{j ϵ D j \neq i} w_{i j} .

In this formulation, a_i is a single marked a-point. D is the set of all points in the entire study area, w_ij is the geographic weight expressing the importance of the j_th point to the i_th marked a-point, x_j is a binary variate value that equals 1 if the j_th point is a marked b-point and equals 0 otherwise. For this analysis, the geographic weights were calculated as Gaussian weights based on the formula,

w_{i j} = \exp [- 0.5 {d_{i j}}^{2} / {d_{i b}}^{2}] .

The variable d_ij² is the square of the Euclidean distance between an individual and another older adult; the variable d_ib² is the square of the bandwidth distance, which is equal to the distance from a_i to the older adult’s 100th nearest neighbor. At distances from a person-condition to another person-condition greater than the bandwidth, the weight formula ratio is greater than 1, and the value of the weight drops off rapidly as distance increases. If the distance from a person-condition to another person-condition equals the bandwidth, the ratio is equal to −1, and the weight is equal to 1 / e^0.5. If the distance from a person-condition to another person-condition equals 0, the ratio equals 0 and the weight equals 1.

The denominator was calculated according to the formula below, where NB is the total number of b-marked points for older adults with a particular chronic condition b in the entire group of older adults, and N is the total number of all older adults, in this case 993:

P (b | a) = N B / (N - 1) .

In a single run of the program, LCLQs were derived for every pairwise combination of chronic conditions, including a chronic condition with itself. The LCLQ recognizes that associations among these combinations of categories may not be symmetric, especially when there are differences in the sizes of the categories. That is, the degree of spatial association of diabetes with another condition such as arthritis might not be equal to the degree of spatial association of arthritis with diabetes. For example, if 20 adults in a population of 100 had diabetes and 80 had arthritis, diabetics could be dispersed among the non-diabetic population such that 80 percent of diabetics’ nearest neighbors were not diabetic, as expected. From the perspective of the 80 individuals with arthritis, however, there are only 20 individuals with diabetes who could be their nearest neighbor so there might be areas with few or no diabetics at all, resulting in higher colocation of non-diabetics than would be expected.

The output table listed the person-condition identifier for each of the 3,335 person-conditions, the unique older adult identifier number, the numeric integer code for the chronic condition of the older adult, and the LCLQs for that person-condition with arthritis, diabetes, heart disease, hypertension, and pulmonary disease. LCLQs greater than 1.0 meant that there were more instances of these other chronic diseases in the neighborhood of the older adult with a particular chronic condition than expected, based on the prevalence in the study population as a whole.

The output for the local CLQs was imported to ArcGIS 10.2. To map spatial associations for a particular combination of chronic conditions, for example, Arthritis-Heart Disease-Pulmonary Disease, all of the output records with LCLQs for those three conditions greater than 1.0 were selected. The selected older persons were then mapped to show whether there were any pockets of concentration of older persons with this combination of conditions.

Results

The five chronic conditions were not prevalent to the same degree among the 993 adults in the study. Hypertension (n = 894) and arthritis (n = 838) were more common than pulmonary disease (n = 545), diabetes (n = 531), and heart disease (n = 527). There were 16 possible groupings of the five chronic conditions of interest among older adults with at least three of the five conditions (Table 1), one with all five conditions, five groups of four of the five conditions, and 10 combinations of three of the five conditions. All of these possible combinations of conditions were observed in the study population.

Patterns of spatial association were investigated for all 10 triads of chronic conditions among individuals with three to five conditions (Table 2). Areas of high spatial association were observed for three triads: Arthritis-Hypertension-Pulmonary Disease (Figure 1), Diabetes-Heart Disease-Hypertension (Figure 2), and Arthritis-Heart Disease-Pulmonary Disease (Figure 3). For the remaining seven triads, no large local areas of high spatial association of the three conditions were found among older adults with three to five chronic conditions. Adults with these combinations of three chronic conditions were distributed in small groups around the state. This kind of observed pattern means that, in every region of the state, the mix of older adults with the three chronic conditions of interest was about the same.

Table 2.

Condition Triad Prevalence Among Adults With Three to Five Conditions.

	Adults (n = 993)
	n
Chronic condition triad
Arthritis-Hypertension-Pulmonary Disease	372
Arthritis-Diabetes-Hypertension	360
Arthritis-Heart Disease-Hypertension	336
Diabetes-Heart Disease-Hypertension	216
Heart Disease-Hypertension-Pulmonary Disease	186
Diabetes-Hypertension-Pulmonary Disease	176
Arthritis-Heart Disease-Pulmonary Disease	172
Arthritis-Diabetes-Heart Disease	162
Arthritis-Diabetes-Pulmonary Disease	154
Diabetes-Heart Disease-Pulmonary Disease	101

Note. The number of older adults does not sum to 993 because individuals with four or five chronic conditions may have more than one chronic condition triad among their four or five conditions.

Figure 1.

High local colocation of arthritis, hypertension, and pulmonary disease among older adults with three or more chronic conditions.

Figure 2.

High local colocation of diabetes, heart disease, and hypertension among older adults with three or more chronic conditions.

Figure 3.

High local colocation of arthritis, heart disease, and pulmonary disease among older adults with three or more chronic conditions.

The full mix of chronic conditions among individuals in the areas of high spatial association for each of the three triads is presented in a series of tables, one for each triad (Tables 3, 4, and 5). Each table summarizes the number of older adults in the areas of high spatial association for the triad by the number of chronic conditions they reported and the prevalence of each of the five conditions. The tables also differentiate older adults with all three conditions from those with only one or two of the conditions in the triad.

Table 3.

Chronic Conditions Among Adults in Areas With Spatial Association of Arthritis-Hypertension-Pulmonary Disease.

	Conditions
	Adults	Arthritis	Hypertension	Pulmonary disease	Diabetes	Heart disease	Sum of all conditions
	n	n	n	n	n	n	N
Adults with all triad conditions
Three conditions	50	50	50	50	0	0	150
Four conditions	22	22	22	22	13	9	88
Five conditions	5	5	5	5	5	5	25
Subtotal	77	77	77	77	18	14	263
Adults without all triad conditions
Three conditions	65	48	55	15	38	39	195
Four conditions	13	10	13	3	13	13	52
Subtotal	78	58	68	18	51	52	247
Total	155	135	145	95	69	66	510

Table 4.

Chronic Conditions Among Adults in Areas With Spatial Association of Diabetes-Heart Disease-Hypertension.

	Conditions
	Adults	Diabetes	Heart disease	Hypertension	Arthritis	Pulmonary disease	Sum of all conditions
	N	n	n	n	n	n	N
Adults with all triad conditions
Three	26	26	26	26	0	0	78
Four	35	35	35	35	24	11	140
Five	24	24	24	24	24	24	120
Subtotal	85	85	85	85	48	35	338
Adults without all triad conditions
Three	178	74	70	157	151	82	534
Four	38	19	20	37	38	38	152
Subtotal	216	93	90	194	189	120	686
Total	301	178	175	279	237	155	1,024

Table 5.

Chronic Conditions Among Adults in Areas With Spatial Association of Arthritis-Heart Disease-Pulmonary Disease.

	Conditions
	Adults	Arthritis	Heart disease	Pulmonary disease	Diabetes	Hypertension	Sum of all conditions
	n	n	n	n	n	n	N
Adults with all triad conditions
Three conditions	8	8	8	8	0	0	24
Four conditions	17	17	17	17	2	15	68
Five conditions	7	7	7	7	7	7	35
Subtotal	32	32	32	32	9	22	127
Adults without all triad conditions
Three conditions	81	68	34	37	35	69	243
Four conditions	21	15	13	14	21	21	84
Subtotal	102	83	47	51	56	90	327
Total	134	115	79	83	65	112	454

Arthritis-Hypertension-Pulmonary Disease

Figure 1 highlights the locations of 155 older adults with three or more chronic conditions, among whose 100 nearest neighbors there was a higher proportion of arthritis, hypertension, and pulmonary disease than expected based on the panel as a whole. There were four areas in the state where older adults had higher than expected prevalence of Arthritis-Hypertension-Pulmonary Disease. The single large area in the central part of the state of stronger spatial association of these conditions was comprised of 131 older adults, and there were three smaller groupings in northern and southern New Jersey.

Of the 993 total adults included, 372 (38%) had Arthritis-Hypertension-Pulmonary Disease among their conditions (Table 2). Of the 155 people in areas of high local associations of these conditions, 77 (50%) had Arthritis-Hypertension-Pulmonary Disease among their three to five conditions (Table 3). Based on a test of single proportion (Fleiss, 1981, p. 13), the proportion is significantly higher for the triad areas than the state as a whole (z = 3.059; p < .05). Within the single large area with 131 older adults, 65 (50%) had all three conditions.

In the sample as a whole, older adults with this triad and exactly three chronic conditions accounted for 27% (186 of 695) of people with exactly three chronic conditions (Table 1) but 65% (50 of 77) of people with exactly three conditions in the area of high spatial association (Table 3). In the large area with 131 older adults, 44% (42 of 96) of people with exactly three chronic conditions had this particular triad of health problems.

As Table 3 shows, 50 of the 77 people (65%) with the Arthritis-Hypertension-Pulmonary Disease triad had only these three conditions. Among people with all three conditions statewide, only 50% had exactly three conditions. Older adults in the areas of high concentration who had all three conditions were less likely to be suffering from additional chronic conditions than in the state as a whole. For those 22 older adults who reported a fourth condition, the condition was more frequently diabetes than heart disease.

About half of the individuals in the areas of high spatial association did not have all three conditions in the triad. Of these 78 adults, only 18 had pulmonary disease. The prevalence of arthritis and hypertension was much higher than pulmonary disease.

Diabetes-Heart Disease-Hypertension

The pattern of spatial association for Diabetes-Heart Disease-Hypertension is different (Figure 2). Among the conditions of the 100 nearest neighbors of the 301 adults highlighted on the map, there was a higher proportion of Diabetes-Heart Disease-Hypertension than we would expect given the frequencies in the state as a whole. Again, there were several areas of strong association, but for this combination, the largest was in the northeast.

Taken as a whole, these 301 adults were almost one third of the 993 people with three or more chronic conditions. Of these 301, 85 (28%) had diabetes, heart disease, and hypertension (Table 4). Yet, among all 993, 216 people had all three of these conditions (Table 2), which in contrast is only 22%. The proportion is significantly higher for the triad areas than the state as a whole (z = 2.655; p < .05). Among the 287 in the largest area of spatial association, 82 (27%) had all three conditions.

In the sample as a whole, older adults with the Diabetes-Heart Disease-Hypertension triad and exactly three chronic conditions account for 8% (55 of 695) of people with exactly three chronic conditions (Table 1). In the areas of high spatial association, 26 out of 204 individuals (13%) with exactly three conditions had Diabetes-Heart Disease-Hypertension (Table 4). In the large area with 287 older adults, 13% (25 of 192) of people with exactly three chronic conditions had this particular triad.

Similar to the Arthritis-Hypertension-Pulmonary Disease areas of high spatial association, there is a higher concentration of people with only the three conditions of Diabetes-Heart Disease-Hypertension in the areas of high spatial association shown in Figure 2. In these areas, 26 out of the 85 individuals (31%) with the three conditions of interest had only the three conditions, while in the state as a whole, 55 out of 216 individuals with all three conditions (25%) had only three chronic conditions. The most prevalent chronic condition in the areas of high spatial association that was not one of the triad conditions was arthritis.

Arthritis-Heart Disease-Pulmonary Disease

The analysis of spatial association of Arthritis-Heart Disease-Pulmonary Disease among older adults with three to five chronic conditions revealed a single large area of high spatial association (Figure 3). This map highlights the locations of 134 older adults among whose 100 nearest neighbors was a higher proportion of arthritis, heart disease, and pulmonary disease than expected. Among these older adults, 32 (24%) had the Arthritis-Heart Disease-Pulmonary Disease triad (Table 5). As shown in Table 2, only 172 (17%) of the 993 adults studied had this chronic disease triad among their three to five chronic conditions. The proportion is significantly higher for the triad area than the state as a whole (z = 1.889; p < .10), but at a lower level than the level observed for the other two triads.

Of the 32, seven had five chronic conditions, 17 had four chronic conditions, and eight had three chronic conditions but all had Arthritis-Heart Disease-Pulmonary Disease in their mix.

Considering only older adults with exactly three chronic conditions, it would be expected that 6% of the individuals in an area within the state would have the Arthritis-Heart Disease-Pulmonary Disease triad because 39 of the 695 older adults in the panel with exactly three chronic conditions had this particular triad (Table 1). In the area of high spatial association, however, 9% (8 of 89) of the older adults with exactly three chronic conditions had this particular triad (Table 5).

Among individuals with all three triad conditions, hypertension was the most prevalent among those with a fourth condition (Table 5). Hypertension was also highly prevalent among the older adults in this area who did not have all three triad conditions. Close to 90% of these individuals reported hypertension while only 55% reported diabetes.

Unlike the areas of high spatial association observed for the other two triads, the proportion of adults with only the three triad conditions (8 of 32 or 25%) was less than the proportion in the state as a whole (55 of 172 or 31%). This triad of conditions is present in adults with four or five chronic conditions to a greater degree than the other two triads.

Discussion

Our results indicated that there is geographic variability in co-occurrence for some multimorbid conditions but not all. Areas with individuals experiencing significantly higher than expected numbers of neighbors with multiple chronic conditions of interest are spatially distinct, with no overlap. This has implications for etiologic research and clinical practice in local communities and at the regional and national levels. Rocca et al. (2014, p. 1345) conclude that there is a need to compare the patterns they observed in the single county they studied “to localized populations in the United States or worldwide to investigate similarities or differences possibly related to environmental, genetic, socioeconomic, or cultural differences.” In addition, there is a need to explore how geographic or community characteristics impact the development and progression of multimorbidity in older adults.

The role of individual behaviors such as smoking, occupational exposures such as to particulates, and environmental conditions including air quality and proximity to major highways could be explored in relation to these apparent patterns of spatial association. In the large cluster of older adults with Arthritis-Heart Disease-Pulmonary Disease conditions (Figure 3), 67% (90 out of 134) had ever tried tobacco, and, of those, 97% (87) had ever used tobacco regularly. In the other cluster of 131 older adults, where Pulmonary Disease was strongly associated with Arthritis-Hypertension (Figure 1), only 52% (69) had ever tried tobacco but 100% of that group reported they had used tobacco regularly. Regular smoking was more commonly found among older adults where Arthritis-Heart Disease-Pulmonary Disease were more strongly spatially associated than expected than among older adults in an adjacent cluster where Arthritis-Hypertension-Pulmonary Disease were more strongly associated than expected. Both of these areas are adjacent to a region of southeastern Pennsylvania where air pollution and climate conditions combine to degrade air quality (Croft & Melendez, 2009).

With regard to Diabetes-Heart Disease-Hypertension, the single large area of strong spatial association is centered on the city of Newark in Essex County, which has a high number and concentration of African Americans and relatively low incomes. Adult diabetes prevalence is highest among African American residents in New Jersey, and among those with lower incomes and less education (Diabetes Control Program, 2005). Comorbidity remains an important research focus, with efforts to compare the quality of treatment for people who have only the index condition with patients who have the disease of interest along with other health problems (de Bruin et al., 2013) and to understand which comorbid conditions share medical care goals with the index condition and which present conflicts in management (Magnan et al., 2015). The method developed here is potentially useful in studying comorbidity. Researchers can select the index chronic condition of interest and look at the local colocation coefficients for the index condition with all other ailments, to identify regions for focused intervention.

For arthritis as an index condition of interest, exercise has been highlighted as an effective non-drug treatment for individuals with osteoarthritis (Hunter & Eckstein, 2009). In both triads including arthritis, pulmonary disease was strongly associated. Yet, chronic pulmonary disease limits exercise (Chin et al., 2013; Cooper, 2001). The areas of strong spatial association of arthritis and pulmonary disease were also areas affected by relatively poor air quality, which would further reduce the opportunities for outdoor physical activity, especially among older adults with pulmonary disease.

The other chronic disease triad of Diabetes-Heart Disease-Hypertension for which areas of strong spatial association were found presents a different picture. A study using Delphi methods concluded that certain forms of heart disease and hypertension were concordant conditions for diabetes as the disease of interest (Magnan et al., 2015). In this area, the chronic conditions share treatment goals with diabetes.

Local spatial statistics can be used effectively to explore the distribution of multiple chronic conditions across the distribution of older adults to uncover areas where more people suffer from particular combinations of chronic conditions than would be expected given the frequencies of these conditions in the population as a whole. The method used here does not require analysts to match exactly on number of chronic conditions, yet it uncovers areas with higher than expected levels of the specific conditions of interest. Research on co-occurrence of chronic conditions in space complements studies summarizing co-occurrence of chronic conditions for communities as a whole (Rocca et al., 2014) and carries implications for practice and future research.

Other methods for joint mapping, such as shared component analysis, seek to derive “global” models for the study area as a whole, require assumptions about data distributions, and have frequently been applied to rate data for areas or districts (Ibáñez-Beroiz, Librero-López, Peiró-Moreno, & Bernal-Delgado, 2011; Knorr-Held & Best, 2000). Local colocation quotients, as modified here, allow researchers to investigate the full range of explicit spatial associations at the level of the individual, even among individuals with slightly different numbers and mixes of conditions. In this analysis, we searched for spatial associations for three chronic conditions at a time among older adults with three, four, or five conditions. There is a degree of similarity in the health status of an older adult with Arthritis-Heart Disease-Pulmonary Disease and another person who has all three of these conditions along with a fourth condition. Such similarities can be further examined to test for possible shared determinants of multimorbidity.

One advantage of spatial analysis implemented in GIS is the support provided for interactive selection and display of observations based on variables such as a fourth and/or fifth chronic condition. Many representations of the large number of chronic disease combinations can be made. More research is needed on methods for visualization of patterns of colocation.

Although this analysis was limited to the 993 people with three or more chronic conditions, the analysis could be extended to all survey participants and other health conditions. Additional conditions, such as various cancers, could also be added. Other conditions would not necessarily have to be chronic diseases. Research in other regions, such as sub-Saharan Africa, has highlighted the need for new frameworks for addressing multiple chronic diseases in the presence of infectious disease epidemics such as HIV/AIDS (Oni et al., 2014)

Self-report of chronic conditions by panel members may provide a different picture of the prevalence of chronic conditions than data obtained directly from medical records. Although almost all (93%) of the 993 older adults included in the study reported having had a physical examination or checkup within the last 12 months, confirmation of the existence and nature of the self-reported chronic problems with medical records would provide a stronger basis of evidence. Data were collected before the Affordable Care Act was enacted, and older adults under 65 years of age were not eligible for Medicare and were not required to have health insurance at the time.

Data used for this analysis are cross-sectional and do not address the development of multiple chronic conditions over time. Older adults reporting chronic health problems such as diabetes, heart disease, and pulmonary disease could have been diagnosed with these conditions when they were children. Future research should focus on expanding the analysis to examine associations among multiple chronic conditions over time.

Despite these limitations, this work adds to the literature by utilizing and expanding current geospatial methodologies and applying them to our understanding of multimorbidity in older adults. Findings demonstrate unique geographic patterns in distributions of multiple chronic conditions across the state of New Jersey. Additional work can be guided by the findings presented here to further our knowledge of state and regional differences in the experience of multimorbidity and inform additional empirical investigations of the factors associated with these differences. Ultimately, this line of inquiry will inform practitioners and policy makers by highlighting areas of public health concern for older adults.

Footnotes

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 author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors thank the UMDNJ/Rowan University School of Osteopathic Medicine whose generous support funded the data collection efforts of the ORANJ BOWL “Ongoing Research on Aging in New Jersey - Bettering Opportunities for Wellness in Life” research panel, and the Swedish Council on Working Life and Social Research FAS dnr 2012-1932, now FORTE, for funding a Visiting Professorship in the Department of Occupational and Environmental Medicine at Lund University during which the methods used in this research were developed.

Author Biographies

Ellen K. Cromley, PhD, is a medical geographer whose research focuses on geographical patterns of health and disease, the location of health services and geographical factors affecting their utilization, and spatial analysis of human health using geographic information systems.

Maureen Wilson-Genderson, PhD, is a quantitative methods psychologist. She serves as consulting statistician for the New Jersey Institute for Successful Aging at the Rowan University School of Osteopathic Medicine.

Allison R. Heid, PhD, has her doctoral degree in human development and family studies with specializations in adult development and aging and intervention science. She currently serves as project director for the New Jersey Institute for Successful Aging at the Rowan University School of Osteopathic Medicine.

Rachel A. Pruchno, PhD, is trained in human development and family studies. Her research focuses on issues related to successful aging. She currently serves as the director of research, university professor, and endowed professor gerontology at the Rowan University School of Osteopathic Medicine’s New Jersey Institute for Successful Aging.

References

Barnet

Mercer

S. W.

Norbury

Watt

Wycke

Guthrie

(2012). Epidemiology of multimorbidity and implications for health care, research, and medical education: A cross-sectional study. The Lancet, 380, 37-43. doi:10.1016/S0149067361260240-2

Blumberg

S. J.

Luke

J. V.

(2007). Coverage bias in traditional telephone surveys of low-income and young adults. Public Opinion Quarterly, 71, 734-749.

Boyd

C. M.

Fortin

(2010). Future of multimorbidity research: How should understanding of multimorbidity inform health system design? Public Health Review, 32, 451-474.

Caughey

G. E.

Vitry

A. I.

Gilbert

A. L.

Roughead

E. E.

(2008). Prevalence of comorbidity of chronic disease in Australia. BMC Public Health, 8, Article 221. doi:10.1186/1471-2458-8-221

Centers for Medicare and Medicaid Services. (2012). Chronic conditions among Medicare beneficiaries (Chartbook: 2012 edition). Retrieved form https://www.cms.gov/research-statistics-data-and-systems/statistics-trends-and-reports/chronic-conditions/downloads/2012chartbook.pdf

Chin

R. C.

Guenette

J. A.

Cheng

Raghavan

Amornputtisathaporn

Cortés-Télles

. . . O’Donnell

D. E.

(2013). Does the respiratory system limit exercise in mild chronic obstructive pulmonary disease? American Journal of Respiratory and Critical Care Medicine, 187, 1315-1323.

Cooper

C. B.

(2001). Exercise in chronic pulmonary disease: Limitations and rehabilitation. Medicine & Science in Sports & Exercise, 33(7 Suppl.), S643-S646.

Croft

P. J.

Melendez

B. V.

(2009). Investigation of the Air Quality Index as related to weather regime. National Weather Digest, 33, 132-153.

Cromley

E. K.

Wilson-Genderson

Christman

Pruchno

R. A.

(2015). Colocation of older adults with successful aging based on objective and subjective measures. Applied Geography, 56, 13-20.

10.

Cromley

R. G.

Hanink

D. M.

Bentley

G. C.

(2014). Geographically weighted colocation quotients: Specification and application. The Professional Geographer, 66, 138-148.

11.

de Bruin

S. R.

van Oostrom

S. H.

Drewes

H. W.

de Jong-van Til

J. T.

Baan

C. A.

Struijs

J. N.

(2013). Quality of diabetes care in Dutch care groups: No differences between diabetes patients with and without co-morbidity. International Journal of Integrated Care, 13, e057.

12.

Diabetes Control Program. (2005). The burden of diabetes in New Jersey: A surveillance report, Section I. Trenton: New Jersey Department of Health and Senior Services.

13.

Feinstein

A. R.

(1970). The pre-therapeutic classification of co-morbidity in chronic disease. Journal of Chronic Diseases, 23, 455-468.

14.

Fleiss

J. L.

(1981). Statistical methods for rates and proportions (2nd ed.). New York, NY: John Wiley.

15.

Ford

E. S.

Croft

J. B.

Posner

S. F.

Goodman

R. A.

Giles

W. H.

(2013). Co-occurrence of leading lifestyle-related chronic conditions among adults in the United States, 2002-2009. Atlanta, GA: Centers for Disease Control and Prevention.

16.

Fortin

Bravo

Hudon

Vanasse

Lapointe

(2005). Prevalence of multimorbidity among adults seen in family practice. Annals of Family Medicine, 3, 223-228.

17.

Freid

V. M.

Bernstein

A. B.

Bush

M.A.

(2012). Multiple chronic conditions among adults aged 45 and over: Trends over the past 10 years. Bethesda, MD: National Center for Health Statistics.

18.

Gilbert

E. W.

(1958). Pioneer maps of health and disease in England. Geographical Journal, 124, 172-183.

19.

Goodman

R. A.

Posner

S. F.

Huang

E. S.

Parekh

A. K.

Koh

H. K.

(2013). Defining and measuring chronic conditions: Imperatives for research, policy, program, and practice. Preventing Chronic Disease, 10, 120239. doi:10.5888/pcd10.120239

20.

Howe

G. M.

(1963). National atlas of disease mortality in the United Kingdom. London, England: T. Nelson.

21.

Hunter

D. J.

Eckstein

(2009). Exercise and osteoarthritis. Journal of Anatomy, 214, 197-207.

22.

Ibáñez-Beroiz

Librero-López

Peiró-Moreno

Bernal-Delgado

(2011). Shared component modelling as an alternative to assess geographical variations in medical practice: Gender inequalities in hospital admissions for chronic diseases. BMC Medical Research Methodology, 11, Article 172. doi:10.1186/1471-2288-11-172.

23.

Knorr-Held

Best

N. G.

(2000). Shared component models for detecting joint and selective clustering of two diseases (Discussion Paper 183). Munich, Germany: Institut für statistic sonderforschungsbereich 386, Ludwig Maximilians Universität. Retrieved from http://epub.ub.uni-muenchen.de/1573/

24.

Kowal

Williams

Wu Fan

Y. J.

Arokiasamy

Chatterji

(2012). Aging, health, and chronic conditions in China and India: A multinational study on global ageing and adult health (SAGE). In Smith

J. P.

Majmundar

(Eds.), Aging in Asia: Findings from new and emerging data initiatives (pp. 415-437). Washington, DC: National Academies Press.

25.

Lehnert

T. S.

Heider

Leicht

Heinrich

Carrieri

Luppa

. . . König

H.-H.

(2011). Health care utilization and costs of elderly persons with multiple chronic conditions. Medical Care Research and Review, 68, 387-420.

26.

LeRoy

Bayliss

Domino

Miller

B. F.

Rust

Gerteis

Miller

(2014). The agency for healthcare research and quality multiple chronic conditions research network: Overview of research contributions and future priorities. Medical Care, 52(3 Suppl.), S15-S22.

27.

Leslie

T. F.

Kronenfeld

B. J.

(2011). The colocation quotient: A new measure of spatial association between categorical subsets of points. Geographical Analysis, 43, 306-326.

28.

Lima

M. G.

Barros

M. B. A.

César

C. L. G.

Goldbaum

Carandina

Ciconelli

R. M.

(2009). Impact of chronic disease on quality of life among the elderly in the state of São Paulo, Brazil: A population-based study. Revista Panamamericana de Salud Pública, 25, 314-321.

29.

Lloyd

C. D.

(2011). Local models for spatial analysis (2nd ed.) Boca Raton, FL: CRC Press.

30.

Lochner

K. A.

Goodman

R. A.

Posner

Parekh

(2013). Multiple chronic conditions among Medicare beneficiaries: State-level variations in prevalence, utilization, and cost, 2011. Medicare & Medicaid Research Review, 3(3). doi:10.5600.mmrr.003.03.b02

31.

Lochner

K. A.

Shoff

C. M.

(2015). County-level variation in prevalence of multiple chronic conditions among Medicare beneficiaries, 2012. Preventing Chronic Disease, 12, 140442. doi:10.5888/pcd12.140442

32.

Magnan

E. M.

Gittelson

Bartels

C. M.

Johnson

H. M.

Pandhi

Jacobs

E. A.

Smith

M. A.

(2015). Establishing chronic condition concordance and discordance with diabetes: A Delphi study. BMC Family Practice, 16, Article 42. doi:10.1186/s12875-015-0253-6

33.

Marengoni

Angleman

Melis

Mangialasche

Karp

Garmen

. . . Fratiglioni

(2011). Aging with multimorbidity: A systematic review of the literature. Ageing Research Reviews, 10, 430-439. doi:10.1016/j.arr.2011.03.003

34.

May

C. R.

Eton

D. T.

Boehmer

Gallacher

Hunt

MacDonald

. . . Shippee

(2014). Rethinking the patient: Using burden of treatment theory to understand the changing dynamics of illness. BMC Health Services Research, 14, Article 281. doi:10.1186/1472-6963-14-281

35.

Meraya

A. M.

Raval

A. D.

Sambamoorthi

(2015). Chronic condition combinations and health care expenditures and out-of-pocket spending burden among adults, Medical Expenditure Panel Survey, 2009 and 2011. Preventing Chronic Disease, 12, 140388. doi:10.5888/pcd12.140388

36.

National Cancer Institute Epidemiology Branch. (1975). Atlas of cancer mortality for U.S. counties, 1950-1969. Bethesda, MD: U.S. Department of Health, Education, and Welfare.

37.

National Center for Chronic Disease Prevention and Health Promotion. (2016). Chronic disease GIS exchange. Division for Heart Disease and Stroke Prevention. Retrieved from http://www.cdc.gov/dhdsp/maps/gixs/

38.

National Center for Health Statistics. (2016). National Health Interview Survey: Questionnaires, datasets, and related documentation. Retrieved from http://www.cdc.gov/nchs/nhis/nhis_questionnaires.htm

39.

Noble

Smith

Mathus

Robson

Greenhalgh

(2012). Feasibility study of geospatial mapping of chronic disease risk to inform public health commissioning. BMJ Open, 2, Article e000711. doi:10.1136/bmjopen-2011-000711

40.

Oni

McGrath

BeLue

Roderick

Colagiuri

May

C. R.

Levitt

N. S.

(2014). Chronic diseases and multi-morbidity: A conceptual modification to the WHO ICCC model for countries in health transition. BMC Public Health, 14, Article 575. doi:10.1186/1471-2458-14-575

41.

Ornstein

S. M.

Nietert

P. J.

Jenkins

R. G.

Litvin

C. B.

(2013). The prevalence of chronic diseases and multimorbidity in primary care practice: A PPRNet report. Journal of the American Board of Family Medicine, 26, 518-524.

42.

Pruchno

R. A.

Wilson-Genderson

Heid

A. R.

(2016). Multiple chronic condition combinations and depression in community-dwelling older adults. The Journals of Gerontology Series A: Biological & Medical Sciences, 71, 910-915.

43.

Pruchno

R. A.

Wilson-Genderson

Rose

Cartwright

(2010). Successful aging: Early influences and contemporary characteristics. The Gerontologist, 50, 821-833.

44.

Rocca

W. A.

Boyd

C. M.

Grossardt

B. R.

Bobo

W. V.

Finney Rutten

L. J.

Roger

V. L.

. . . Sauver

J. L.

(2014). Prevalence of multimorbidity in a geographically defined American population: Patterns by age, sex, and race/ethnicity. Mayo Clinic Proceedings, 89, 1336-1349.

45.

Rushton

Armstrong

M. P.

Gittler

Greene

B. R.

Pavlik

C. E.

West

M. M.

Zimmerman

D. L.

(2008). Geocoding health data: The use of geographic codes in cancer prevention and control, research, and practice. Boca Raton, FL: CRC Press.

46.

Shannon

G. W.

(1981). Disease mapping and early theories of yellow fever. The Professional Geographer, 33, 221-227.

47.

United Nations Department of Economic and Social Affairs, Population Division. (2013). World population prospects: The 2012 revision, highlights and advance tables (Working Paper No. ESA/P/WP.228). Retrieved from http://esa.un.org/unpd/wpp/publications/Files/WPP2012_HIGHLIGHTS.pdf

48.

van den Akker

Buntinx

Knottnerus

J. A.

(1996). Comorbidity or multimorbidity: What’s in a name? A review of the literature. European Journal of General Practice, 2, 65-70.

49.

van den Akker

Buntinx

Metsemakers

J. F. M.

Roos

Knottnerus

J. A.

(1998). Multimorbidity in general practice: Prevalence, incidence, and determinants of co-occurring chronic and recurrent diseases. Journal of Clinical Epidemiology, 51, 367-375.

50.

Ward

B. W.

Schiller

J. S.

Goodman

R. A.

(2014). Multiple chronic conditions among US adults: A 2012 update. Preventing Chronic Disease, 11, 130389. doi:10.5888/pcd11.130389

51.

Wolff

J. L.

Starfield

Anderson

(2002). Prevalence, expenditures, and complications of multiple chronic conditions in the elderly. Archives of Internal Medicine, 162, 2269-2276.