The Design Utility of Rural Hospital Community Health Needs Assessments

Introduction

Rural communities, defined in the United States as geographically distant and sparsely populated areas, continue to face barriers accessing quality healthcare (Coughlin et al., 2019; Douthit et al., 2015). Limitations including inadequate or no insurance coverage (Bolin et al., 2015; Day, 2019; Goldfield & Fuller, 2019), lack of community resources, minimal local healthcare provider training, constraints in transportation (Iglehart, 2018), and absence of necessary service lines (Bolin et al., 2015; Brems et al., 2006). In addition to the lack of these resources critical for quality care delivery, rural communities are at risk of reduced access to care due to the increase of rural hospital closures for the last decade (Kaufman et al., 2016). Hospital closures between 2010–2019 resulted in over 175,000 (1%) of regional populations without hospital access within 15 minutes driving time (McCarthy et al., 2021). Hospital closures in rural settings leave many without life-saving care, with 50% of rural residents already experiencing travel times of 30 minutes or more (Akinlotan et al., 2023). Vehicle-based ground transportation time to access care increases in rural areas of the country often leading to delayed medical care (Wolfe et al., 2020). The overall lack of resources available within and external to critical access hospitals (CAHs) can be a contributing factor in closures as the facility does not have the means to remain agile and respond to the needs of an evolving population.

Background

Critical Access Hospital (CAH) is a designation given to certain hospitals in rural communities that meet a set of criteria as defined by the Centers for Medicare & Medicaid Services (Schoenbaum, 2011). Criteria to meet CAH designation includes: 25 or fewer acute care beds, located further than 35 miles from another hospital, provides 24/7 emergency care, and maintains an average length of stay for acute care patients of 96 hours or less (Critical Access Hospitals (CAHs), 2024). CAHs are critical for providing care in rural settings where access to larger hospitals is costly or transportation prohibitive. Additionally, CAHs provide life-saving emergent care to stabilize patients and transfer, when needed, to larger care facilities (Seright & Winters, 2015).

Many CAHs were constructed between the 1950s and 1970s with federal funds appropriated by the Hill-Burton Hospital Survey and Construction Act, Pub. L. 79–725, 60 Stat. 1040, enacted July 13, 1946. Communities often struggle in their strategic planning of renovations and updates, resulting in these facilities aging faster due to deferred maintenance than other hospitals nationwide (Call, 2024). The extensive changes to equipment, technologies, service lines and care complexity have led most facilities to a point where new construction is more cost-efficient to revitalize the built environment to better meet community needs for both outpatient and inpatient services (Call, 2024). The United States Department of Agriculture (USDA) Rural Development fund includes direct loans (7 CFR Part 1942, Subpart A) and grants (7 CFR Part 3570, Subpart A) to help finance construction costs and upgrades in furniture, fixtures, and equipment, but generally only funds a portion of the costs with the rest raised through bonds or other fundraising. It is then tasked to the CAH leadership to develop a strong business case to the community that the new construction or renovation creates necessary avenues by which to support health-positive behaviors and economic growth. The high visibility of these healthcare projects and community investment requires acknowledgement and reflection of resident feedback in the construction as well as yielded benefits of design considerations proposed.

Community Health Needs Assessment

Prior to any new construction planning and operational programming, design teams seek information about the community served by the healthcare facility. CAHs are unique in that they are required to collate an extensive report, conducted by survey, on the healthcare utilization behaviors, attitudes surrounding health, and motivations of community members to engage with the healthcare facility. Known as a Community Health Needs Assessment (CHNA), this tool is used to gather information on the self-reported health and health-related goals of a specific population, understand the resources available to meet these goals, and describe the community’s needs to achieve the quality of life associated with their meaning of health as a construct (Ravaghi et al., 2023). CHNAs have historically been used to help guide and create community health improvement (CHI) plans by healthcare facilities to be responsive to community changes and adapt to evolving care delivery considerations, ranging from local policy to federal mandates (Stoto et al., 2019).

However, CHNAs are largely static reports which minimizes the ability to exchange knowledge of the changing health needs in rural communities and may not be raised as a resource to outside design teams when they initially engage in ideation sessions with healthcare leadership or community forums. In small population settings such as rural communities, CHNAs lend a larger sample of feedback from individuals around the catchment area otherwise not easily accessible by a design team. CHNAs have the potential to support the collective healthcare and design industry understanding of needs in siloed, rural communities by giving community members a voice in the steering of larger initiatives and establishing mutual goals between the healthcare facility and community.

Purpose

For participatory design to be effective, understanding the broader health needs driving the design change is necessary. Perceptions of health across individual and community levels often drive the decision to access care at a facility as well as the types of care needed for wellness across the lifespan (primary, hospital, specialty). Analysis of the CHNA provides a picture of the experience receiving care as a rural resident in the targeted service area and its surrounding population centers. CHNAs are required by law for healthcare facilities to remain designated as a CAH and retain their respective tax status, making their availability both consistent in approach and predictable as a source of data to inform design decision-making. The primary objective of this study was to examine current healthcare utilization patterns of rural residents via secondary analysis of a CHNA for a rural CAH in northeastern Montana currently engaged in built environment research for rural healthcare facilities. The secondary objective of this study was to identify relationships between rural resident perceptions of individual and community health, indicators associated with likelihood to access care, and community insights which may guide design decision-making for future construction projects.

This study’s short-term goal was to describe the infrastructure needs and community healthcare services necessary to sustain care delivery at this CAH such that informed decision-making may occur related to plans for renovation of the hospital and/or plans for lines of services congruent to resources available. The long-term goal is to leverage identified areas of opportunity for heightened efficiency and revenue to support a profitable, sustainable healthcare system for the residents of the CAH’s service area and its surrounding areas. This longer-term initiative includes updated infrastructure, comprehensive personnel recruitment and retainment plans, and deepened partnership with community leaders to empower the CAH to be a responsive, adaptable system of care.

Methods

Guiding Framework

The Rural Community Participatory Design Framework (RCPDF) is a design framework by which active community members participate in informing design choices in the healthcare-built environment (Johnson & Hellem, 2024). The RCPDF overlaps common phases of a design project – Ideation+Alignment; Planning; Construction; Occupancy; Evaluation; and Mutual Goal-Setting – with known contextual cultural considerations during these phases. An example is the need to establish familiarity (perceptions of being welcomed or unwelcomed) during the Occupancy project phase, which can be supported through culturally responsive transition and activation planning (Johnson & Hellem, 2024). Another example in this cyclical framework is the appraisal of health behaviors and health beliefs during the Mutual Goal-Setting phase to segue into Ideation+Alignment canvassing of resources and networks associated with the behaviors and beliefs (Johnson & Hellem, 2024). This type of design framework goes beyond community participatory research and involves the community in the actual design process. The key to this design framework is to not merely encourage or facilitate dialogue but fosters growth of both the community members and the design team to occur across project phases. Through this interconnectivity, the user group, client, and design team will have a more holistic view of community and the community’s needs.

This process to achieve a holistic view of community needs can be facilitated by a regional Community Design Centers (CDCs) through action research, in which the significant transformations of the user group and the design group can be promoted (Mehrhoff, 1999). The Montana Community Design Center, in the School of Architecture at Montana State University, is one such organization that utilizes action research. The tradition of Montana State University’s CDC originates in the seminal use cases as referenced by Mehrhoff (1999), including a community design organization, the Minnesota Design Team, whose values-centered approach highlights the importance of making connections:

The design professionals who work with a community are also making connections among themselves and to the data about the host community. The ancient axiom “Physician, heal thyself” applies equally to community design as a healing process. Just as community members explore the nature of their community in a number of ways, sharing insights and ideas from a variety of backgrounds and perspectives, community designers such as members of a Minnesota Design Team also undertake the demanding process of building their own community (p. 101).

The Montana Community Design Centers follows a similar ethos and reflects the same intent, which serves as a cornerstone for the RCPDF. Through this framework, the user is empowered to be a designer, and the designer becomes the user. By using existing resources such as CHNA data gathered by community members, the design team is afforded a head start in creating community context. During the Ideation+Alignment phase, secondary analysis of CHNA data can be used within the RCPDF to appraise existing and identify non-existing resources or networks that will help or hinder project progression (Johnson & Hellem, 2024). This allows for alignment of costs, materials, labor, and other core components of project management. The CHNA is an investment of time and financial resources for the healthcare facility and thus significant to leverage during project management as common ground of community voices and an additional lens of healthcare utilization patterns. Viewing the community as the client via the RCPDF, the design team delivers project outcomes culturally congruent to the community’s meaning of health foundational to the rural way of life and pervade the meaning of health and healthcare. The framework is cyclical in nature, meaning that once the design process is complete, the evaluation process begins again. Continuous evaluation of data is necessary to allow the system to be adaptive to the community’s changing needs.

Project Design

This study is a retrospective quantitative, descriptive, secondary analysis of a CHNA dataset for a CAH to identify areas of community perspective convergence and divergence from the observed findings in the hospital environment which may affect healthcare access and utilization of local services. The Office of Rural Health for Montana conducted a CHNA based on pre-determined lines of questioning and survey structure, which was mailed to residents in the service county and adjacent areas that are included in the catchment zone of the CAH. The initial analysis performed by the Office of Rural Health primarily included descriptive statistics and qualitative interpretation, with an ultimate deliverable to the CAH leadership of an executive summary of the findings and recommendations based on the results. The Office of Rural Health completed a comparison between the most recent two CHNA datasets (2021 and 2024) to determine patterns of differing responses among community members completing the surveys. The present secondary analysis was driven by the pre-established comparison.

Ethical Considerations

This study was approved through the Montana State University Institutional Review Board (# 2023–851-Expedited). As part of the regulatory review, initial permission was obtained by the CAH leadership to request access to the raw data and collaborate with the Office of Rural Health to glean additional potential insights that otherwise may be not readily visible in the traditional CHNA report. In alignment with federal reporting guidelines related to low census populations, generalized categories were established a priori to support anonymity of residents, particularly with survey items that had low response (i.e., less than or equal to 10 responses). Given the methodological focus on this analysis, proper nouns and identifying information within the survey questions and responses were removed given heightened familiarity among the population. Once the CHNA dataset was prepared by the Office of Rural Health, it was transferred by secure, encrypted file management within the University system to a shared repository only accessible by personnel identified in the Institutional Review Board protocol.

Setting

Of the 67 hospitals established in Montana, 55 are considered critical access and are facing a closure rate of 1 in 4 facilities (Rural hospitals at risk of closing, 2024; Top hospitals in Montana by net patient revenue, 2024). The surveyed CAH is a six-bed Critical Access Hospital with an attached Rural Health Clinic (RHC) located in the northeastern plains of Montana. The CAH has a service area of 5,100 square miles and provides medical services to the immediate communities’ residents totaling 4,192 people and the adjacent three counties. The CAH’s county is designated by the US Department of Health and Human Services as a health professional shortage area due to the low ratio of primary care and specialty providers per resident, which also qualifies this region as a frontier county (i.e., low density of persons per square mile as well as distance and travel time to service/market). The CAH’s primary service area includes its local town, two villages, and four population centers as designated by the United States Census Bureau (2024). The CAH’s key services include emergency medical services, inpatient and outpatient care, skilled nursing and swing bed capabilities, laboratory, radiology, physical therapy/occupational therapy, and a durable medical equipment supplier. The family health clinic includes services such as primary care, telehealth, and outreach clinic providers in the area (e.g., ophthalmology) once or twice monthly. Currently, the CAH has an estimated 300 total patient days per year and a revenue of nearly $11 million. There is another CAH within 70 miles to the West and a slightly larger CAH within 90 miles to the South. The closest Level I facility (highest level of trauma care) is 210 miles, which under ideal conditions via ground transport is over three hours of transit time across rural, two-lane highways or roadways, with limited or extremely limited services between townships and small municipalities.

Sample and Recruitment

The CAH provided an aggregated list of outpatient and inpatient admissions and information regarding service area zip codes. The included zip codes with the greatest number of admissions were selected to be included in the survey. A random list of 800 residents was then selected with the assistance of the Social Data Collection and Analysis Services (i.e., a fee-for-service core facility at Montana State University Bozeman). This core facility enables the collection, cleaning, and presentation of high-quality data for researchers using a variety of social and behavioral methods. Residence was stratified in the initial sample selection so that each area would be represented in proportion to the overall served population and the proportion of past admissions. This stratification enabled a balance of both populations concentrated in clustered towns and those in more frontier settings, such as ranches or farms with fewer than six persons per square mile. Additionally, survey completion was targeted for even distribution as possible to male and female representing populations given the disproportional response rate among females traditional to both rural setting and survey distribution method given their high involvement in healthcare behaviors in family structures (Koolwal, 2021; Wu, Zhao, & Fils-Aime, 2022). Surveys were mailed to participants and could be returned via mail or completed online via a provided web link. Proportional representation of the geographic areas was the most important factor considered in distribution. As such, age was not considered when proportionally sampling.

Measures

Forty-two questions were included, of which four questions were demographic measures. Items of measurement included questions such as perceptions of personal health (11 questions), perceptions of community health (9 questions), barriers to receiving care (8 questions), and quality of care (1 question) (Table 1). Question response items were asked on a binary, Likert scale, all that apply, single select, or to select the top three.

Table 1.

Community Health Needs Assessment Survey Questions and Measures

Survey Question Number	Question Text	Variable	Response Options	Variable Level
	Model Responses
15	In the past three years, have you or a household member seen a primary healthcare provider such as a family physician, physician assistant or nurse practitioner for healthcare services?	Primary Care	Yes/No	Binary
18	In the past three years, has anyone in your household received care in a hospital? (i.e. hospitalized overnight, day surgery, obstetrical care, rehabilitation, radiology or emergency care)	Hospitalist Care	Yes/No	Binary
21	In the past three years, have you or a household member seen a healthcare specialist (other than your primary care provider/family doctor) for healthcare services?	Specialist Care	Yes/No	Binary
	Model Predictors
1	How would you rate the general health of our community?	Community Health	Likert	5a
2	How would you rate your own personal health?	Personal Health	Likert	5 a
5	How do you rate your knowledge of the health services that are available through the [Critical Access Hospital]?	Knowledge of Health Services	Likert	4b
35	How well do you feel your health insurance covers your healthcare costs?	Insurance Coverage	Likert	4b
39	Where do you currently live?	Residence	Single Select	8c
40	What is your gender?	Gender	Single Select	4
41	What age range represents you?	Age	Single Select	8
42	What is your employment status?	Employment	Single Select	8
	Descriptive Analyses
13	In the past three years, was there a time when you or a member of your household thought you needed healthcare services but did NOT get or delayed getting medical services?	Delay in Care Access	Yes/No	Binary
16	Where was the primary healthcare provider located?	Primary Care Location	Single Select	12c
17	Why did you select the primary care provider you are currently seeing?	Primary Care Provider Reason	Multiple Choice	13d
19	Which hospital does your household use MOST for hospital care?	Hospital Location	Single Select	6c
20	Thinking about the hospital you were at most frequently, what were the three most important reasons for selecting that hospital?	Hospital Resources	Multiple Choice	13e
22	Where was the healthcare specialist seen?	Specialist Location	Multiple Choice	9c
23	What type of healthcare specialist was seen?	Specialist Type	Multiple Choice	34f
34	What type of health insurance covers the majority of your household’s medical expenses?	Insurance Type	Single Select	11g
36	Have you ever missed a medical appointment due to a transportation issue?	Transport	Yes/No	Binary
37	If you do not have health insurance, why?	No Insurance Reason	Multiple Choice	5h

Note. 42 total questions were used on the survey. The selected above questions showed relevance for analyses.

^aLikert scale selections: Very healthy, Healthy, Somewhat healthy, Unhealthy, and Very Unhealthy

^bLikert scale selections: Excellent, Good, Fair, and Poor

^cResponse options hidden to maintain anonymity

^dSelections: appointment availability, reputation for quality, recommended by family/friends, and referred by physician

^eSelections: required by insurance plan, prior experience with the hospital, and referred by physician

^fSelections OB/GYN, Neurologist, Audiologist, Urologist, Oncologist, Gastroenterologist, and General Surgeon

^gSelections: employer sponsored, government sponsored coverage programs, health insurance marketplace, or other

^hSelections: cannot afford to pay, employer does not offer, choose not to, too complicated to sign up/re-apply, or other

Focus area

For this secondary analysis, primary items of interest included: perceptions of health, both personal and community, insurance coverage, and barriers to care. Information relating to knowledge of services available was also of interest. With an aging population, knowledge of services available can also be a barrier due to communication channels. Rural communities will sometimes receive specialists that come for single or multi-day clinic appointments. These appointments can happen once a month, once a quarter, or even more infrequently. Alerting the community of availability to see traveling specialists can present a unique challenge in knowing which medium to communicate through (e.g., magazine ads vs. Facebook ads).

Data Analysis

Data were de-identified and stored in password protected digital storage. Due to the data coding of the initial analysis, some items had to be recoded to fit a binary coding, such as 1 for “yes” and 0 for “no.” Secondary descriptive, inferential, and predictive analyses were performed using Minitab statistical software (V22, Minitab LLC, State College, PA). Descriptive analyses were used to understand respondent’s characteristics and perceptions of aspects of care including access and barriers. Chi-square association tests were performed to identify univariate trends amongst perceptions and respondent characteristics.

Leveraging the Rural Community Participatory Design Framework, stakeholder interviews via a community advisory board and observational assessments of the CAH built environment were conducted prior to analyses of the CHNA data (Johnson & Hellem, 2024). This exploratory, contextual inquiry approach identified key areas of interest among community members and hospital leadership to isolate in the CHNA data for comparative analysis and determination of clinical utility of CHNA responses towards physical infrastructure decision-making. Indicators identified included perceptions of the facility to meet health needs, utilization behaviors for current and potential service lines, and proposed CAH offerings based on population level demographics (community composition). Based on these a priori indicators with that highest degree of clinical utility that influenced design of the healthcare-built environment, binary logistic regression models were constructed to determine the relationship between accessing primary, hospital, and specialty healthcare in the last three years and eight predictors. For each of the three binary responses (yes accessed care in the last 3 years), four predictors were included that were Likert-rating of respondent’s perception of community health, personal health, health insurance coverage, and knowledge of health services available. Four additional demographic predictors were included to explain any differences in accessing care based on residence, gender, age, and employment.

An exploratory Classification and Regression Trees (CART^® Classification, V22, Minitab, LLC, State College, PA) was used to model the same three binary responses with respect to the eight predictors. CART prediction was used to illustrate the patterns and relationships with the binomial responses and specified predictors (Breiman et al., 1984). As an alternative to standard parametric methods, these three CART classification models were primarily used to gain insight into relative variable importance and identify groups in the data with desirable characteristics. Each CART classification model’s nodes, misclassification cost, relative variable importance, confusion matrix, receiver operating characteristic (ROC) curve, gain chart, and lift chart were reviewed prior to acceptance (Breiman et al., 1984; Gordon, 1984). Given the small sample sizes, CART was used as a proof-of-concept to analyze CHNA data for future more robust analyses across multiple CAHs. Given the imbalanced relative frequencies for the binary response, all binary logistic regression and CART classification models used “Yes accessed care within the last three years” (1) as the event of interest. The significance level was set at α=0.05 for all statistical analyses. All analyses were reviewed for practical and clinical significance with rural healthcare experts to maximize interpretability for community stakeholders.

Results

2024 CHNA Response Rate and Respondent Demographics

One hundred and seventy-five surveys were returned out of 800. Of those 800 surveys, 50 were returned undeliverable resulting in a 23.3% response rate (175/800). Over half of the responses were received from the primary city (58.8%, Table 2), which is the main geographic area served by the CAH. The remaining responses were from smaller surrounding communities (35.1%). While age was not a consideration when sampling, the results of the age categories of respondents are consistent with the aging population of other rural communities (Cohen & Greaney, 2023). The Office of Rural Health performed a similar CHNA in 2021. The present research team did not analyze the data; however, access was provided to the final report to compare descriptive statistics of the sample. The largest category of respondents were ages 66–75 in the 2021 and 2024 CHNA. In 2021, respondents ages 66–75 made up 27.1% of all responses. This percentage increased to 32.9% (n=57) in 2024.

Table 2.

Community Health Needs Assessment Survey Respondent Demographics

Demographic Question	Category Total Respondent Count	Sub-Category Count and Frequency
	n	n	%
Residence	174
Within Critical Access Hospital Population Center		102	58.62
Within 40 Miles of Population Center		61	35.06
Greater than 60 Miles or ‘Other Location’		11	6.32
Age (Years) n=173	173
18–35		17	9.83
36–45		20	11.56
46–55		14	8.09
56–65		37	21.39
66–75		57	32.95
76–86+		28	16.18
Gender	171
Female		112	65.50
Male		59	34.50
Employment Status	173
Full-Time		64	37.00
Part-Time		14	8.09
Retired		67	38.73
Othera		28	16.18

^aIncluded students, those respondents collecting disability, and respondents self-identified as unemployed

Of the respondents that chose to identify their gender (n=171), 34.5% were male (n=59), and 65.5% were female (n=112). Non-binary and other were additional options for gender identification. No respondents identified with these categories. The high response rate of females as compared to males is not surprising as women tend to be the primary healthcare decision makers of their household (Matoff-Stepp et al., 2014). Respondents were also asked to indicate their employment status. Unsurprisingly, given the age of the population, the majority of respondents indicated they were retired (n=67). If employed, most respondents indicated they worked full time (n=64).

Several of the respondents reported not accessing any type of care in the last 3 years. Specifically, 6% had not accessed any type of primary care, 37% hospital-based, and 22% specialty care in 36 months. As this survey was conducted in 2024, there is an overlap with the COVID-19 pandemic of 2021. For those that only accessed primary care services, most were older between 56 and 75 years old (55%), were retired and had good insurance coverage. For those that did not access any primary care services (6%), most had fair to poor insurance coverage, tended to be younger (18–35 years old) and had another type of employment. For those that did access primary care and at least one other service, most accessed hospital-based care (78% vs. 22%, respectively). While there were respondents who did not access care, a vast majority reported accessing one to multiple healthcare services in the last three years. General perceptions of those that did access care thought their personal and community health was at least somewhat healthy, had good to excellent health insurance coverage, and felt their knowledge of health services in the community was at least fair to good. Over half of the respondents lived locally to the CAH, were female and worked full-time.

Accessed Primary Care in the Last Three Years

Nearly all respondents accessed primary healthcare services in the last three years (94.12%), but only 11.8% were seen in the CAH’s attached rural health clinic (RHC) or local provider. Most reported their primary care services were in cities located 130+ miles (30.4%) and 200+ miles (16.2%) from the CAH’s located city. Both alternate facilities have ground transportation times greater than 2 hours one-way under optimal conditions. Most selected their primary care services to be closest to home (38.5%) and prior experience with the selected clinic (39.1%). Nearly 40% of respondents reported a delay in seeking healthcare. Respondents most often delayed care because they could not get an appointment (26%), felt it cost too much (26%), concerned about the quality of medical providers (26%), or took too long to wait for an appointment (24%). There were additional delays due to concerns over provider choice, respect and when offices were open (>10% each). Transportation, distance, childcare, and employment were not cited as primary sources for delays in care (<10% each). Interestingly, only 13.5% reported missing any type of medical appointment due to transportation issues or delays.

Using forward selection, binary logistic regression determined that insurance coverage and age were significantly associated with being seen by a primary care provider (Table 4). Compared with a perception of poor insurance coverage, both perceived good (p=0.018) and excellent (p=0.036) coverage were significantly associated with a higher odd of accessing primary care in the last 3 years. A majority of respondents stated they had government-sponsored insurance (41.5%, Medicare/Medicaid/Veteran Affairs) followed by employer-sponsored insurance (23%). Less than 5% of respondents stated not having insurance of any kind primarily due to the cost of medical insurance (63%). Similarly, those 36–55 years old (p=0.025) and 56–75 years old (p=0.003) were significantly more likely to access primary care services compared to their younger counterparts (18–35 years old).

Table 4.

Logistic Regression Model of Primary Care Patient Encounters in Last Three Years

Predictor Variable Category and Response Rangea	Coefficient	Standard Error	Test Statistic	P-Value	Odds Ratio 95% Confidence Interval
Constant	−0.694	0.926	−0.75	0.453	-
Insurance Coverage b
Goodc	2.36	1.00	2.36 (1)	0.018***	10.6 (1.49, 75.28)
Excellentc	2.68	1.27	2.1 (1)	0.036***	14.5 (1.19, 176.3)
Age (Years) d
36–55e	2.92	1.31	2.24 (1)	0.025***	18.6 (1.44, 240.5)
56–75e	3.18	1.07	2.97 (1)	0.003***	24.1 (2.95, 196.1)
76+e	2.27	1.30	1.75 (1)	0.080	9.67 (0.76, 122.8)

^***Denotes significance of p<0.05

^aForward selection was used at a keep alpha=0.25

^bAnalysis of Variance Wald χ²=7.28 with 2 degrees of freedom, p-value = 0.026

^cReference response value was ‘Fair to Poor’

^dAnalysis of Variance Wald χ²=10.05 with 3 degrees of freedom, p-value = 0.018

^eReference group was 18–35 years old

Accessed Hospital Care in the Last 3 Years

Most respondents accessed hospital services in the last three years (63.4%) with 26.9% being seen in the local CAH. However, hospital-based services were also accessed at the closest two CAHs requiring one-way commutes of 90 miles (18.5%) and 71 miles (23.2%), respectively. Most sought care at alternate facilities due to privacy/confidentiality concerns (49%), closer to their residence (35%), emergent situation (31.5%), and the hospital’s reputation for quality care (31.5%). Using binary logistic regression, personal health, community health, age and employment status were forward selected into the model for hospital-based care (Supplemental Table 1). The factors of personal health, community health, age and employment status, while included at an alpha of 0.25, failed to reach the level of significance. Those who perceived the community as somewhat healthy, retired, or worked part-time were slightly less likely to access hospital-based care (p = 0.087, 0.104, 0.095, respectively). Conversely, those aged between 36–55 years old were slightly more likely to access hospital care (p=0.100, Odds Ratio = 3.79). The large service area and the adjacent available two CAHs in nearby counties resulted in increased variability for usage at the studied CAH.

Accessed Specialist Care in the Last 3 years

Most respondents accessed specialty care services in the last three years (77.7%) with only 16.0% being seen in the local CAH. A majority accessed specialty services at the closest Level I facility, which has the largest number of specialties in the state. Some specialists were accessed at the closest two CAHs (29.8% and 22.1%, respectively). Respondents most frequently reported being seen by a dentist (28.5%), orthopedic surgeon (26.2%) optometrist (23.85%), and cardiologist (21.5%). None (0%) of the respondents indicated that that had been seen in the last 3 years by a dietician, geriatrician, psychiatrist, social worker or substance abuse counselor. An additional 13 specialties were seen by 10 or more respondents. Only 21% combined accessed any type of therapy service (physical and occupational) and audiology, which occupy a significant portion of the outpatient clinic and inpatient services.

Personal health, community health, insurance coverage and gender were forward selected into the binary logistic regression model for specialty care (Supplemental Table 2). Those who perceived their personal health as healthy to very healthy were significantly more likely to access specialty care (p=0.049). Conversely, those who perceived the community as healthy to very healthy, were significantly less likely to access specialty care (p=0.046). Of note, both groups are contrasted with those that perceived their own health and the community’s health as unhealthy. Those with excellent insurance coverage were significantly more likely to access specialists for their care (p=0.094). Males were slightly more likely to access specialty care than females (p=0.14).

Exploratory Predictive CART Analysis

While all eight predictors were included in the forward selection process, knowledge of health services and zip code of residence both failed to meet the selection alpha of 0.25 for all three response models. Moreover, gender and employment status were only marginally significant for a single model. Despite the relatively low sample sizes, CART models were used to identify classifications among the categorical predictors and binary responses which otherwise did not emerge using regression approaches (Table 5). Perceptions of community health, employment status, insurance coverage and age were the top three predictors amongst the three CART models for accessing primary, hospital and specialty care. These strong relationships identified in the CART model were likely not identified using the regression models due to the complexity and underlying patterns difficult for standard statistical techniques to identify. Those accessing specialty care showed a diverse range of perceptions and groupings, which led to the most complex 9-node decision tree heavily branched towards excellent insurance coverage and perceptions of their community’s and personal health. Uniquely, employment status, residence, and gender provided no relative importance in accessing specialty care in Montana, likely due to the lack of specialty care within the state. While an exploratory predictive analyses, the relatively simple 4-node models for accessing primary and hospital-based care provide an excellent decision tree for education to the community and reducing barriers by healthcare providers in the area. A strong focus on communication and educating personal and community health with access to at least good health insurance coverage may enable better usage of local resources for primary and hospital care.

Table 5.

CART Classification Model Comparison Across Responses

Response	Primary Care n=160	Hospital Care n=109	Specialty Care n=132
Terminal Nodesa	4	4	4
Relative Misclassification Costb	0.4063	0.9032	0.8640
Relative Variable Importance c
Community Health	100%	34.9%	55.2%
Employment	87.7%	66.6%	-
Insurance	47.5%	49.9%	100%
Age	43.1%	66.1%	31.5%
Knowledge of Health Services	28.8%	1.90%	15.9%
Residence	16.2%	26.0%	-
Personal Health	9.30%	100%	44.6%
Gender	2.40%	0.30%	-
Sensitivity d
Trainingf	78.1%	71.6%	69.7%
Testg	79.4%	54.1%	58.3%
Specificity e
Trainingf	100%	57.1%	76.3%
Testg	80.0%	55.6%	55.3%
Type I Error
Trainingf	0.0%	42.90%	23.70%
Testg	20.0%	44.40%	44.70%
Type II Error
Trainingf	21.9%	28.40%	30.30%
Testg	21.0%	45.90%	41.70%

^aA terminal node is a final node that cannot be split further and is used to make predictions

^bRelative misclassification cost for the model relative to a trivial classifier that organizes all observations into the most frequent class

^cPredictors in order of their effect on model improvement from all the basic functions for a predictor. Standardizes the important values for ease of interpretation

^dTrue positive rate

^eTrue negative rate

^fPortion of the dataset that was entered into the CART model for pattern identification

^gPortion of the dataset used to test the CART prediction model

Example Impact on CAH Proposed Renovation

A key finding of this study was the low or unreportable resident response rate of accessing physical/occupational therapy or audiology, yet these two specialty offerings comprise a significant footprint, approximately 6,000 (20%) of the approximately 30,000 combined square footage of outpatient clinic and hospital. At the time of initial contextual observation by the study team, the healthcare leadership had discussed expanding the physical/occupational therapy (PT/OT) space, which is contradictory to the reported utilization by this study sample. A targeted environmental assessment of the shelled space proposed for the expanded PT/OT was found to be inherently acoustically dampened on the basement level of the facility and located next to the only elevator bay. The aging population in the community and need for primary and prevention services warrant consideration to use this space as noted in the CHNA for supporting healthy aging in place.

As the community served by the CAH continues to age, there is ever-growing challenge of providing care to a population with corresponding medical care needs and complex case management (Cohen & Greaney, 2023; Wu et al., 2022). Older adults in rural communities are often socioeconomically disadvantaged and face increased barriers to receiving primary healthcare (Ford et al., 2016). Availability of primary and preventative care services has been shown to reduce emergency department use for non-emergent care needs, reducing costs to a facility and allows for the emergency department to remain clear for appropriate cases (Timmins et al., 2020). As the shelled space is directly located a floor below the emergency department and adjacent to the existing PT/OT, there is opportunity to co-create a preventative care area with exercise equipment, self-serve health monitoring equipment (e.g., blood pressure), and exam space for rotating primary and specialty care providers specific to older adult needs. This conversion is substantiated by CHNA quantitative data and qualitative resident feedback provided as part of the CHNA survey. The healthcare leadership, empowered by this secondary analysis, has agreed to move forward with design team collaboration on ideating this proposed concept and seeking funds to support its development. The research team provided the CAH with an executive summary of findings and recommendations to relay to the design team during goal-setting and initial alignment (RCPDF phases).

Discussion

This secondary analysis of a rural CHNA aligned with existing research surrounding the perceptions of community health being a strong predictor of the likelihood to seek medical treatment (Harju et al., 2006). Previous research in rural North Carolina described rural residents as ‘ambivalent’ with ‘irregular health habits’ to seek care when there were fewer community health-related behaviors or cultural norms surrounding trust of hospitals (Harju et al., 2006, p. 361). In this study, it was observed that rural residents demonstrated this ambivalence through contrasting responses between their perceptions of personal health and the health of the community. The similar findings between this study in Montana and North Carolina demonstrate a cultural connection across geographically differing rural communities where resource availability and trust in hospitals to provide care access influence care seeking behaviors.

As shown in this study as with other research comparing community needs in both rural and urban settings as a systematic review, ability to pay for services is a major determinant to seek care (Harju et al., 2006; Cyr et al., 2019). Insurance coverage data in the CHNA strongly predicted if a rural individual was more likely to access care in general as well as specialty care, and this was also corroborated in the systematic review of rural and urban research when availability and perception of care was a recurring theme (Cyr et al., 2019). As both rural and urban settings seek to leverage telehealth to reduce costs and travel burden, there is still disparity in home digital access, Internet, and insurance coverage. Among residents most vulnerable to increased morbidity and mortality, such as shown in one study with American Indians, Alaska Natives, Hispanics, and non-Hispanic Blacks (Curtis et al., 2021). Compoundingly, both rural and urban communities have noted challenges with shared information exchange and communication across clustered healthcare facilities or practices that may be frequented by a local population. Ineffective communication leads to redundancy in diagnostics and delays in care plan harmonization across primary and specialty providers, which allocates time away from value-added patient care and moreso on coordination that burdens both rural and urban residents (Cyr et al., 2019).

Knowledge of local CAH services did not prove to be a significant factor in healthcare access behaviors across any of the models. Based on this secondary analysis of this CHNA, a respondent’s insurance coverage status and type may drive decision-making particularly when seeking primary and preventative care. However, insurance coverage and type of insurance was not significantly related to being hospitalized in the last three years, which may be in part given the emergent or urgent situation where care is necessitated regardless of insurance status. As the question did not specify the reason for a hospital visit (emergent vs. non-emergent), a visit to the hospital could also be indicative of the upward trend in rural communities of using emergency services for primary care reasons (Baskin et al., 2015).

As is the case across the United States, rural residents are aging at a faster rate than urban centers (Cohen & Greaney, 2023). Aging populations, such as the population identified in this study, require anticipatory considerations to healthcare and general services offered in order to ‘age in place.’(Cohen & Greaney, 2023). Aging in place largely rests on cultural and environmental factors that promote wellness and decrease isolation among older adults, thus affording improved health monitoring and resource allocation to support food security, social involvement in the community, and safety (Cohen & Greaney, 2023). However, older adults have experienced constrained services in their rural communities that leave gaps in necessary care attributable to changes in local economies as young adults depart from low-resource (rural, frontier) settings to metropolitan and micropolitan centers (Thiede et al., 2017). Environments conducive to healthy aging extend beyond the hospital walls, as evidenced by this secondary analysis, to perceptual implications of the community at-large being ‘healthy’ or ‘unhealthy’ if key adjacencies are not available. Common adjacent resource examples include grocery stores, areas for physical exercise, community centers, and public transportation (Klann et al., 2019). This CHNA analysis highlighted the absence of respondent documentation or discussion of these common resources known to the community and incongruence of knowledge related to resources correlating with healthcare access. Individuals identified sporadic hours of operation for high school physical exercise facilities but did not recall other resources such as the city hall weight room or outdoor swimming pool local to the hospital. While health and health behaviors were assessed, key adjacencies were not captured in a manner that would be brought forth as significant findings, whereas design teams commonly will examine key adjacencies as part of site analysis during the Ideation + Alignment phase noted in the RCPDF. Healthcare leadership may then be made aware of the importance of appraising knowledge of key adjacencies, particularly for older adults, and the impact on the perception of health and positive health behaviors.

A secondary analysis of CHNAs can provide additional feedback to CAHs on questions that are more relevant to the needs of the community and perceptions of quality that inform future design choices while acknowledging the differing degrees of adaptability with local resources. While the traditional approach of descriptive statistical CHNA reports provide useful information, diving deeper into data analysis and looking at the associative results between questions allows for a more robust interpretation of the data with a higher degree of clinical utility. For example, an association between insurance status and access to primary care services may indicate absence of local facilities or differences in respondent behaviors, norms, or values associated with primary or preventative care.

Implications and Recommendations for Rural Healthcare Design

While contextual observation and qualitative interviews are the ultimate standard for community participatory design, the expense of gathering enough data in interviews can be cost prohibitive (Rupert et al., 2017). As CHNAs are a fixed requirement, they can be an anticipatory bridging tool and resource between design team and the rural community to rapidly assess the fit of a project’s scope and budget. Qualitative research surrounding physical design needs of CAHs has been conducted, such as one study in North Carolina, where staff provided insights via surveys and interviews surrounding the needs of the CAH with focus on the physical facility, patient/provider satisfaction, and safety (Zamani, Joy, & Gobel, 2023). However, broader community voice was not incorporated to validate or contradict the reports from 21 individual staff members (Zamani, Joy, & Gobel, 2023). The incorporation of the CHNA for the CAH adds additional points of discussion, increased rigor to survey generation by design teams, and generates higher information power, or richness of insights, from focus groups conducted across project phases (Morse, 2020). CHNAs are tools in starting conversations between design teams and community members through their own words and perceptions. Ultimately, the early integration of community voice from pre-existing resources (CHNAs) enhanced by culturally informed ideation sessions achieve more meaningfulness to a sustainable design approach given local resources to motivate residents to engage in health-related behaviors.

CHNA data analysis findings should not act as a complete substitute for qualitative interviews. Rather, synergistic interpretation of focus group data collection should be used to either validate convergence or divergence from CHNA analysis findings. Divergence from CHNA findings should be further investigated before design changes are finalized. This may be indicative of inadequate population sampling in the CHNA process, or imbalanced focus group composition of stakeholders. An example is the heavy emphasis on expanding physical/occupational therapy, which is gaining attention as a shortfall in reimbursement for rural areas compounded by absence of trained providers (Small Rural Hospital and Clinic Finance 101 Guide, 2021). However, the CHNA does not depict sufficient attention to this service line in comparison to absent service lines which would meet other meaningful design goals, such as inclusion of obstetrical care and veteran’s health, which are both known healthcare access shortages in this region of Montana, particularly among American Indian communities which border the CAH (Charlton et al., 2016). Design teams may then be responsive to these identified service lines that are under-developed or not yet included in the operational planning of the proposed space in the CAH, demonstrating the importance of leveraging the CHNA as a tool to promote conversation with the client and evidence in the design process.

CHNAs are helpful in that the data is widely available and gives community voice perspective through utilization of retrospective data acquisition. In its current format, however, CHNAs are isolated by system or state, with little affordance to readily collate data for large representations of geographical regions. Future recommendations include collation of CHNA data among design firms as part of market surveillance to understand regional and state-wide patterns of healthcare utilization, which in turn permits anticipatory prospective patterns over time. This would allow for more collaboration amongst healthcare facilities that are responsive in their decision-making related to service line inclusions and the implications for built environment evaluations and engagement with community leaders.

Direct community engagement has limitations and barriers—primarily time and cost—however, the concept of self-build, in which the user group becomes involved with the construction, has grown in the United Kingdom and has interesting implications for the United States (Jenkins et al., 2010). New construction, renovations, and rehabilitative hospital projects have a direct connection to economic development of rural towns, which largely remain in persistent poverty and experience dwindling opportunities for employment where the succession planning of ranches and farms remains constrained (Rhoades, Whitacre, & Davis, 2022; Stermer, 2024). Many critical access facilities in the United States run at a financial deficit which can mean that prior to closing altogether, certain branches of care are shut down or mergers take place to conserve budget (Carroll et al., 2023). Reinvestment in rural areas through community-involved construction of health sector facilities, as employees, co-designers, and builders, has been further validated in a large sample of cities in China where presence of providers and proximity to care through construction of facilities were deemed instrumental indicators of rural prosperity (Du & Jiao, 2023). Since there is high visibility by the community, and financial backing by the community is inherently needed, low capital projects may be more desirable since they can be completed by a local, less formal approach, helping to ameliorate the lack of resources, time, or budget. Design teams can then provide a unique value proposition by including CHNAs as part of their engagement with a community through tailored ideation, anticipated use of local labor, and co-development during Planning (RCPDF phase) that allows for trust to be created between insiders and outsiders from that community.

As other research has reported, useful indicators of community health and wellbeing are subjectively selected and often lack evidence-base to adequately inform materials selection, construction schedule, and thematic architectural elements crucial for community representation (Lach et al., 2022). This study and the included framework concurred with global research of community assessments design utility in that the approach here was grounded in the community’s capabilities and their knowledge base related to design and definition of health (Ravaghi et al., 2023). Mutual Goal-Setting (RCPDF phase) includes connecting health behaviors and beliefs to operational programming recommendations of the design team. Capabilities in operational planning of a healthcare project may include level of technological integration, ability to communicate via phone/computer service, and the presence of local economy which substantiates the value proposition for a service line or facility (Ravaghi et al., 2023). Broadband availability and cell tower counts are reduced in rural areas compared to urban environments, with tower counts as low as 10 per 1,000 km² in rural Indiana versus 165 per 1,000 km² in Chicago (Zhang et al., 2022). Supplemental to CHNAs, which ask questions related to telehealth use and other digital means of care delivery, design teams may support community decision-making during operational programming and goal-setting by inquiring the connectivity resources which may determine the degree of digital integration in the facility (Zhang et al., 2022).

Limitations and Future Research

Rural populations are known to be quantitatively small, racially/ethnically homogeneous, and geographically distanced. Given these established characteristics of rurality, the effect size in this study was small though many of the associations and models were significant. As this study utilized an exploratory CHNA, collation of additional years of CHNA data may allow for more robust data interpretation. Questions included on the CHNA were not augmented for this study and lacked validated measures for certain inquiries, such as asking respondents about their level of depression. Questions permitted openness with interpretability by respondents, which in turn widened the response variability and reduced the generalizability afforded from narrowing response options. This widened response potential created many categories of responses that were too few to report.

To address limitations noted here in future research, the investigators will collaborate with the Montana Community Design Center, the Office of Rural Health, and partnering CAHs to determine changes in approach with survey lines of questioning to optimize use of results across different design projects. With Community Design Centers advocating for social services and community projects, the Montana Community Design Center will lead the initiative with the engagement of architecture and nursing students. The core mentality of this initiative will be conducted through an empathic approach and as noted earlier under the RCPDF, the designers’ (including students) and the user groups’ roles will not remain static and flip back-and-forth. It is the intent that these two groups will experience transformational change where the user is empowered to be a designer and the designer a user. Through this attitude and approach through empathy, surveys and interviews may have more efficacy and better response rate. As the response rate was 25.6% (188 residents) for the 2021 CHNA, there is opportunity to reimagine the sampling approach and methods to expand the inclusion of additional voices from the surrounding communities to increase the response rate and to support engagement through the Montana Community Design Center, as well as the students. This has included raising the CHNA as an important instrument of change management and utility for design teams with the state healthcare association stakeholders.

The next phase of this research will include an industry-wide national survey across architecture, engineering, construction, and integrated design groups/firms to appraise the RCPDF and use of CHNAs during design projects as a resource and tool for ideating, aligning, and familiarizing the team with the local community. This will be completed in collaboration with the Montana Community Design Center in order to employ participatory design methods to explore and test best practices in the field. The results of this study will provide scaffolding for developing highly interpretable CHNA reports for design team use and how these reports are best presented across the project cycle through the RCPDF framework. This future research will build upon published qualitative findings from Zamani, Joy, and Gobel (2023) which identified environmental, cultural, and perceptual factors that contributed to the identification of key areas of hospital renovation and rehabilitative design. Additionally, the next phase will expand on the topics of power; interdisciplinary design and engagement; and making meaningful connections with low-resource communities vulnerable to poor health outcomes.

Additionally, the predictive modeling was exploratory and having yielded positive results can now be applied to a much larger sample size to understand regional healthcare paradigms using CART and other more advanced artificial intelligence modeling. Future work will include a larger sample size of CHNA data across all Montana CAHs to predict response values for new observations. RCPDF is an iterative process, requiring a constant feedback loop between CAHs and the communities of interest during renovations, rehabilitations of care delivery spaces, and complete redesign projects. Through analysis of CHNAs, interpretation of findings, focus groups, and improvement of survey design, CAH leadership and design teams can utilize community voice as derived from CHNAs to inform project design and execution.

Conclusions

This project, which consisted of a secondary data analysis of a community health needs assessment for a small CAH in rural Montana, was conducted to expand our understanding alignment between observable built environment findings which may affect care access, and behaviors compared to resident survey responses. Through this study, it is suggested that visits to a primary care facility, specialty clinic, or the hospital have certain associated predictors, such as insurance, perceptions of personal health, or perceptions of community health. Novel to this analysis is the finding that respondent knowledge of service offerings in the area was not a significant predictor of engaging with CAH services.

As CHNAs are consistently conducted and public sources of information, there is a high degree of utility for design teams during ideation and project conceptualization to gain insights from a larger sample of a geographically expansive population outside stakeholders immediately local to a facility or those identifying as employees. CHNAs may also be used to guide healthcare facilities and systems in low-resource settings to develop post-occupancy evaluations and performance metrics which are congruent to the built environment in which care is delivered. The RCPDF was validated in its adaptability to incorporate retrospective sources of information to guide phases of a design project and integrate external sources of community data to lend additional insights, which may guide client decision-making. Future research will be pursued to include longitudinal review of CHNAs compared to augmentations in the built environment and healthcare service lines of CAHs to understand regional patterns of resident healthcare seeking and access behaviors.

Table 3.

Descriptive Analyses of Predictor Variable by Responses (Event Only)

Predictor Variable Category and Response Range	Primary Care	Hospital Care	Specialist Care
	n (%)	n (%)	n (%)
Personal Health	159	108	132
Very Unhealthy to Unhealthy	10 (6.29)	8 (7.41)	8 (6.06)
Somewhat Healthy	52 (32.7)	42 (38.9)	46 (34.9)
Very Healthy to Healthy	97 (61.1)	58 (53.7)	78 (59.1)
Community Health	158	108	131
Very Unhealthy to Unhealthy	14 (8.86)	13 (12.1)	13 (9.92)
Somewhat Healthy	101 (63.9)	63 (58.3)	87 (66.4)
Very Healthy to Healthy	43 (27.2)	32 (29.6)	31 (23.7)
Insurance Coverage	154	107	128
Fair to Poor	31 (20.1)	25 (23.4)	28 (21.9)
Good	81 (52.6)	53 (49.5)	61 (47.7)
Excellent	42 (27.3)	29 (27.1)	39 (30.5)
Knowledge of Health Services	154	107	128
Fair to Poor	31 (20.1)	25 (23.4)	28 (21.9)
Good	81 (25.6)	53 (49.5)	61 (47.7)
Excellent	42 (27.3)	29 (27.1)	39 (30.5)
Residence	159	109	132
Critical Access Hospital Population Center	93 (58.5)	60 (55.1)	53 (40.2)
Gender	156	107	129
Female	104 (66.7)	70 (65.4)	82 (63.6)
Male	52 (33.3)	37 (34.6)	47 (36.4)
Age (Years)	158	108	132
18–35	13 (8.23)	12 (11.1)	11 (8.33)
36–55	32 (20.3)	28 (25.9)	28 (21.2)
56–75	90 (57.0)	53 (49.1)	73 (55.3)
76+	23 (14.6)	15 (13.9)	20 (15.2)
Employment Status	158	108	132
Full-Time	60 (38.0)	47 (43.5)	49 (37.1)
Part-Time	13 (8.23)	NRb	12 (9.1)
Retired	62 (39.2)	35 (32.4)	50 (37.9)
Othera	23 (14.6)	NRb	21 (15.9)

^aIncluded students, those respondents collecting disability, and respondents self-identified as unemployed

^bNot reportable (NR) due to low sample sizes for response selection

Implications for Practice.

• Use of the Community Health Needs Assessment (CHNA) data supports evidence-based design discussions with the healthcare facility’s leadership pertaining to infrastructure and operational planning needs in accordance with essential services (primary care, specialty care) identified by the community members.

• The variance in personal and community perspectives of health and healthcare seeking behaviors collated in CHNAs can alert design teams to cultural and contextual nuances influencing significance of the facility in the local community for collective health needs, further informing later stakeholder ideation sessions, mutual goal-setting, and user-centered participatory design methods.

• CHNAs, when collated from many healthcare systems and facilities, may be compared internally by design teams to describe regional landscapes of service line or operational gaps and identify points of consensus across community ecosystems to anticipate design priorities, such as technological integration, demographic-specific space needs, and construction logistics.