Social Determinant of Health Documentation Trends and Their Association with Emergency Department Admissions

Abstract

Research has shown that health outcomes are significantly driven by patient’s social and economic needs and environment, commonly referred to as the social determinants of health (SDoH). Standardized documentation of social and economic needs in healthcare are underutilized. This study examines the prevalence of documented social and economic needs (Z-codes) in a nationwide inpatient database and the association with emergency department (ED) admissions. Multivariate logistic regression was used to assess the effect of social and economic Z-codes on hospital admission through the ED. Payer source, gender, age at admission, comorbidity count, and median ZIP code income quartile covariates were included in the logistic regression analyses. Patients with documented social and economic Z-codes were significantly more likely to be admitted through the ED than those without documented social and economic needs, after adjusting for covariates. Standardized and widespread collection of these valuable Z-codes within EHR systems or administrative claims databases can help with targeted resource allocation to alleviate possible barriers to care and mitigate ED utilization.

Introduction

Inpatient care accounts for approximately 31 percent of health expenditures in America, and nearly 50 percent of all hospitalizations originate in the emergency department (ED)^1,2. This high rate of emergency service utilization may be due to patient preferences, access and availability of primary care services, or physician referral patterns for inpatient admission¹. According to data from the National Emergency Department Survey, ED visits increased by 18.4 percent, and ED hospitalizations increased by 6.8 percent between 2006 and 2014³.

To combat the rapid growth of healthcare spending, value-based payment programs have focused on decreasing non-urgent and inappropriate ED visits and hospitalizations through more holistic care. The Patient Protection and Affordable Care Act spawned new care delivery and payment models, including Accountable Care Organizations and State Innovation Models that have emphasized the identification and incorporation of patients’ social and economic needs into clinical care. Research has supported these strategies, demonstrating that health outcomes are significantly driven by an individual’s social and economic status and environment, commonly referred to as the social determinants of health (SDoH)^4-6.

The World Health Organization defines the SDoH as the conditions in which individuals work, live, worship, and age⁷. Social determinants can have direct and indirect effects on health^8-13. However, SDoH indicators are rarely captured in electronic health record (EHR) systems^14-16. Making SDoH available as specific risk factors and providing targeted interventions could positively influence care through data exchange among clinicians and social service professionals¹⁷. We posit that collecting SDoH data within EHR systems can help clinical teams support upstream medical and social care issues to decrease utilization of ED visits.

Integrating SDoH data into EHR systems requires a mechanism to capture these data in a standardized form. The International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM) includes a set of social, economic, and behavioral factors that impact how care is received, known as V-codes. This set of codes were expanded and refined in the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM). ICD-10-CM contains a subset of reason codes known as Z-codes to capture “factors that influence health status and contact with health services¹⁸.” These code sets represent an opportunity to collect standardized data on social and economic needs in any health care setting and to be leveraged for referrals to social services from community-based organizations.

While becoming more accepted due to recent research and investment in SDoH, social care has long been a tangential, often neglected, component to standard medical care^19-21. So too is the collection of social and economic needs data in EHR systems²². Documentation rates for social and economic needs are low, likely due to the lack of universally accepted standards and lack of reimbursement for the services that require these specific indicators. It was estimated that only 1.6% of discharges leveraged ICD-9-CM V-codes before the inception of ICD-10-CM Z-codes16. However, more refined Z-codes are also sparsely documented. Centers for Medicare and Medicaid Services (CMS) reported that Z-codes were documented for only 1.4 percent of the 33.7 million fee-for-service Medicare beneficiaries in 2017²³.

Even with efforts underway to integrate Z-codes into payment formulas and clinical practice, the need for the proper and universal documentation of these factors must become clearer^24-26. Efforts by the Institute of Medicine and the National Association of Community Health Centers made preliminary mappings between SDoH screening items and ICD-10-CM codes^17,27. Furthermore, studies are beginning to emerge that link social and economic needs documented in EHR systems to utilization patterns. For instance, a recent study found a correlation between a subset of documented social and economic needs and future hospitalization and ED utilization²⁸. In application, risk assessment tools, including machine learning models, that combine social and clinical information have proven more valuable in assessing ED utilization⁹.

Though Z-codes have the potential to help hospitals track SDoH, it is unclear how often they are utilized and their relationship with the consumption of healthcare services. Given the high cost of ED visits, and their often-implied lack of preventive care, understanding SDoH and ED admissions could have both clinical and financial implications. This study examines the prevalence of SDoH Z-codes in a nationwide database and the association with ED admissions, and variation in association by subgroup. To our knowledge, this study is the first to analyze the correlation between documented SDoH Z-codes and admission through the ED using a national sample. This work is also noteworthy because it examines patient attributes that are available for any inpatient admission, requiring no prior utilization knowledge on the patient and can be easily reproduced by healthcare facilities.

Methods

This research was a retrospective case-control study using inpatient admission data from the Health Care Utilization Project (HCUP) to understand the relationship between social and economic needs with admission through the ED. HCUP is a family of software tools and databases developed and sponsored by the Agency of Healthcare Research and Quality. Historical data collection began in 1988 and are “derived from administrative data and contain encounter-level, clinical and nonclinical information including all-listed diagnoses and procedures, discharge status, patient demographics, and charges for all patients, regardless of payer²⁹.” These data are available to purchase for research and reporting purposes. The HCUP houses the Nationwide Readmissions Database (NRD) that contains discharge data from 27 states, accounting for 56.6 percent of all US hospitalizations. These data provide a link to connect hospital admissions without revealing dates of service or location to meet privacy guidelines. Admission attributes included in the HCUP NRD include length-of-stay, ICD-10-CM diagnoses, procedure codes, patient demographics, and payment related elements. The data are de-identified and exclude the 16 identifiable variables that necessitate IRB approval for access³⁰. Because of the de-identified nature of the data, this study is not human subjects research and is considered exempt from IRB according to the policy of the National Institutes of Health Office of Human Subjects Research.

For this study, data from the HCUP NRD were extracted and stored in a distributed hive 7-node cluster and accessed and accessed using SparkR for cleaning and analyses. HCUP NRD is not trackable across calendar years. Therefore, admissions were limited to 2016, allowing for all relevant features to be extracted using IDC-10-CM diagnosis codes, which was adopted by the majority of healthcare providers in October 2015. Only non-elective admissions were considered for analyses since planned inpatient services do not admit through the ED. Additionally, this study focused on admissions for patients that were aged 18 years or older.

Admissions defined as part of the case cohort contained a SDoH Z-code in one of the 35 ICD-10-CM fields for a given admission. SDoH Z-code families of interest, as highlighted by the American Hospital Association, can be found in Table 1³¹. Case patients must have had a diagnosis code beginning with the first three characters of the Z-code families of interest. For instance, patients with a documented Z60.2 (problems living alone) or Z60.4 (social exclusion and rejection) would both receive an indicator for Z60 (problems related to social environment). Due to the low prevalence of documented SDoH Z-codes, a representative, random sample was drawn from the remainder of adult, non-elective admissions to form the control cohort.

Table 1.

Median Annual Income Quartile Thresholds (2016)30

Quartile	Median Annual Income Thresholds
0 – 25th Percentile	$1 - $42,999
26th – 50th Percentile	$43,000 - $53,999
51st – 75th Percentile	$54,000 - $70,999
76th – 100th Percentile	$71,000+

Health status was defined using the Charlson Comorbidity Index (CCI), quantifying the number of conditions that appeared on an admission via the 35 available ICD-10-CM diagnosis codes. Each patient’s CCI was categorized as 0, 1, or 2 or more conditions^32,33. While patient ZIP codes were not provided for privacy reasons, median household income quartiles were available. HCUP NRD suppressed median household income quartile for ZIP codes with populations below a threshold or states with only a single ZIP in a particular quartile to further protect patient confidentiality. Quartile thresholds used to define this attribute for 2016 can be found in Table 1.

Demographic and socioeconomic attributes were compared between case and control cohorts using the Chi-squared test of independence. Multivariate logistic regression was used to assess the effect of SDoH Z-codes on ED admissions. Payer source, gender, CCI, age at admission, and median ZIP income quartile covariates were included in the logistic regression analysis. All data compiling, cleaning, and analyses were conducted using R (version 3.5.2) with α = 0.05. The ‘ICD’ R package was used to define CCI³⁴.

Results

Out of nearly 16 million adult admissions in the HCUP NRD for 2016, a total of 304,146 unique, non-elective admissions had a documented SDoH Z-code of interest. A randomly selected 724,460 admissions did not have a documented SDoH Z-code. Figure 1 depicts the logical formation of the case and control groups from the original HCUP NRD.

Figure 1.

Sampling Schematic

The frequency distribution of SDoH Z-codes of interest is shown in Table 2. SDoH Z-codes related to housing and economic circumstances (Z59 family) were documented most frequently (63 percent). The Z57 family of SDoH Z-codes, dealing with exposure to occupational risk factors, had the most substantial documentation rate differences between patients admitted through the ED and those not admitted through the ED, 87.1 percent and 12.9 percent, respectively. Problems related to certain psychosocial circumstances (Z64 family) was the only SDoH Z-code family with a higher percentage of patients not admitted through the ED (66.3 percent). While each Z-Code family contained a different number of unique codes, the correlation between the number of codes in each family and the total number of documented codes was insignificant (r = 0.064).

Table 2.

Frequency Distribution of SDoH Z-codes

		Total	Not Admitted Through ED (%)	Admitted Through ED (%)
Any Z-Code		304,146	59,026 (19.4)	245,120 (80.6)
Z-Code Family	AHA Description31
Z55 – Problems related to education and literacy (7 distinct codes)	Illiteracy, schooling unavailable, underachievement in a school, educational maladjustment and discord with teachers and classmates.	1,978	897 (45.3)	1,081 (54.7)
Z56 – Problems related to employment and unemployment (12 distinct codes)	Unemployment, change of job, threat of job loss, stressful work schedule, discord with boss and workmates, uncongenial work environment, sexual harassment on the job, and military deployment status	48,511	13,563 (28.0)	34,948 (72.0)
Z57 – Occupational exposure to risk factors (12 distinct codes)	Occupational exposure to noise, radiation, dust, environmental tobacco smoke, toxic agents in agriculture, toxic agents in other industries, extreme temperature, and vibration.	2,245	290 (12.9)	1,955 (87.1)
Z59 – Problems related to housing and economic circumstances (10 distinct codes)	Homelessness, inadequate housing, discord with neighbors, lodgers and landlord, problems related to living in residential institutions, lack of adequate food and safe drinking water, extreme poverty, low income, insufficient social insurance and welfare support.	191,664	32,062 (16.7)	159,602 (83.3)
Z60 – Problems related to social environment (7 distinct codes)	Adjustment to life-cycle transitions, living alone, acculturation difficulty, social exclusion and rejection, target of adverse discrimination and persecution.	32,168	6,204 (19.3)	25,964 (80.7)
Z62 – Problems related to upbringing (24 distinct codes)	Inadequate parental supervision and control, parental overprotection, upbringing away from parents, child in welfare custody, institutional upbringing, hostility towards and scapegoating of child, inappropriate excessive parental pressure, personal history of abuse in childhood, personal history of neglect in childhood, Z62.819 Personal history of unspecified abuse in childhood, Parent-child conflict, and sibling rivalry.	24,620	7,806 (31.7)	16,814 (68.3)
Z-Code Family	AHA Description31
Z62 – Problems related to upbringing (24 distinct codes)	Inadequate parental supervision and control, parental overprotection, upbringing away from parents, child in welfare custody, institutional upbringing, hostility towards and scapegoating of child, inappropriate excessive parental pressure, personal history of abuse in childhood, personal history of neglect in childhood, Z62.819 Personal history of unspecified abuse in childhood, Parent-child conflict, and sibling rivalry.	24,620	7,806 (31.7)	16,814 (68.3)
Z63 – Other problems related to primary support group, including family circumstances (14 distinct codes)	Absence of family member, disappearance and death of family member, disruption of family by separation and divorce, dependent relative needing care at home, stressful life events affecting family and household, stress on family due to return of family member from military deployment, alcoholism and drug addiction in family	37,531	10,772 (28.7)	26,759 (71.3)
Z64 – Problems related to certain psychosocial circumstances (3 distinct codes)	Unwanted pregnancy, multiparity, and discord with counselors.	594	394 (66.3)	200 (33.7)
Z65 – Problems related to other psychosocial circumstances (8 distinct codes)	Conviction in civil and criminal proceedings without imprisonment, imprisonment and other incarceration, release from prison, other legal circumstances, victim of crime and terrorism, and exposure to disaster, war and other hostilities.	15,728	4,898 (31.1)	10,830 (68.9)

Characteristics of both the case and control cohorts are shown in Table 3. Admissions with documented SDoH Z-codes were more likely to be financed through Medicaid, directly from the patient (i.e., self-pay), or “other.” Patients with documented SDoH Z-codes were more likely to be younger, male, and reside in ZIP codes with median income in the lowest quartile. Patients with documented SDoH Z-codes were more likely to have a lower CCI.

Table 3.

Frequency Statistics of Case vs. Control Cohorts

Attribute	Without Documented SDoH Z-Code (%) (N = 724,460)	With Documented SDoH Z-Code (%) (N = 304,146)	p-value
Gender			< 0.001
Female	401,352 (55.4)	116,066 (38.2)
Male	323,108 (44.6)	188,080 (61.8)
Mean Age (std. dev.)	59.9 (20.1)	47.4 (16.7)	< 0.001
CCI			< 0.001
0	197,118 (27.2)	134,043 (44.1)
1	130,383 (18.0)	72,665 (23.9)
>= 2	396,795 (54.8)	97,438 (32.0)
Median ZIP Income Quartile			< 0.001
1st to 25th percentile	240,747 (33.6)	113,791 (38.8)
26th to 50th percentile	155,307 (21.7)	69,949 (23.8)
51st to 75th percentile	162,041 (22.6)	62,328 (21.2)
76th to 100th percentile	158,004 (22.1)	47,483 (16.2)
Payer			< 0.001
Medicaid	134,840 (18.6)	135,966 (44.7)
Medicare	372,569 (51.4)	83,100 (27.3)
Private	176,261 (24.3)	37,837 (12.4)
Self-Pay	24,618 (3.4)	29,030 (9.5)
Other	16,172 (2.2)	18,213 (6.0)

Case patients were significantly more likely to be admitted through the ED than control patients, after adjusting for sociodemographic attributes, payer source, and CCI. Table 4 provides the odds of particular subgroups being admitted through the ED compared to the respective reference group. A 95 % confidence interval (CI) is reported to quantify the statistical significance of each relationship.

Table 4.

Logistic Regression Results

Attribute	Admission through ED Odds Ratio (95% CI)	p-value
Social & Economic Need	1.23 (1.21 - 1.24)	< 0.001
Gender (Ref = Male)
Female	0.72 (0.71 - 0.73)	< 0.001
Age	1.02 (1.02 - 1.03)	< 0.001
CCI (Ref = 0)
1	1.52 (1.50 - 1.54)	< 0.001
>= 2	1.68 (1.66 - 1.71)	< 0.001
Zip Income (Ref = 0-25th percentile)
26th to 50th percentile	0.91 (0.89 - 0.92)	< 0.001
51st to 75th percentile	1.04 (1.03 - 1.06)	< 0.001
76th to 100th percentile	1.06 (1.04 - 1.08)	< 0.001
Payer (Ref = Medicaid)
Medicare	0.91 (0.90 - 0.93)	< 0.001
Private	0.77 (0.76 - 0.78)	< 0.001
Self-Pay	1.66 (1.62 - 1.70)	< 0.001
Other	0.71 (0.69 – 0.73)	< 0.001

The adjusted odds of being admitted through the ED for case patients were 1.23 times that of control patients (95% CI = (1.21 – 1.24)). Female patients were less likely to be admitted through the ED (OR = 0.72, 95% CI = (0.71 – 0.73)). The odds of ED admission increased by two percent with each additional year increase in patient age (95% CI = (1.02 – 1.03)). For patients with an assigned CCI value of ‘1’ or ‘>=2’, the odds of an ED admission increased by 52 percent and 68 percent, respectively. Patients living in poorer ZIP codes were less likely to be admitted through the ED, with patients in the highest quartile being six percent more likely to admit through the ED compared to the lowest quartile (95% CI = (1.04 - 1.08)). Admissions funded through the patient (i.e., self-pay) were 66 percent more likely to be admitted through the ED compared to Medicaid admissions (95% CI = (1.62 - 1.70)). All other payer sources were less likely to be admitted through the ED.

Discussion

While social needs assessments in the clinical setting are becoming more commonplace, back-end data efforts to extract information from more universal systems is lacking. Health system and payer specific interpretations of standard taxonomies are continuing to emerge; therefore, the opportunity to act soon at a macro-level is imperative³⁵. As more healthcare organizations utilize these valuable code sets to inform decision-support efforts, the easier it will be for interoperability of information between medical care and social service organizations.

In the HCUP NRD, less than two percent of adults with a non-elective admission had a documented SDoH Z-code. The majority of documented SDoH Z-codes pertained to inadequate housing and living circumstances. This set of codes includes indicators of welfare support. Further analysis uncovered that 53 percent of patients with a Z59 family documented code were financed through Medicaid. The Z64 family of codes was the only grouping to be less likely to be admitted through the ED. This family describes psychosocial issues, including unwanted pregnancy. Ninety-seven percent of patients with a documented code from the Z64 family were female, and the mean age of this group was 32.7 years, far younger than other groups. Occupational risk factor codes, the Z57 family, had the highest ED admission rate. This may be indicative of injuries and risks incurred on the job that warrant hospitalization.

This study demonstrated that even when controlling for other factors, SDoH Z-codes were correlated with ED admission. By comparison, having at least one documented SDoH Z-code increased the odds of admission through the ED by nearly the same amount as having two or more comorbidities. The association of documented SDoH Z-codes with ED admission may be indicative of screening tools and protocols executed explicitly in the ED versus during an inpatient stay or at discharge. Furthermore, the association of social and economic needs and ED admissions may be an extension of well-studied link between lack of access to preventive care and utilization of low-value services.

Additional refinement and adoption of standardized codes sets should aid in producing specific and relevant information on social and economic needs of patients on a broader scale. To help, a draft of the International Classification of Diseases, Eleventh Revision (ICD-11) reveals enhancement in these specific code sets³⁶. Institutions such as CMS and the Department of Health and Human Services should be actively evaluating the potential of these code sets to help identify and stratify the risk of patients for various barriers to care.

Other non-clinical factors were significantly associated with ED admission. These factors were evaluated to determine if the well-studied associations remained consistent with the incorporation of social and economic needs. Payer, gender, age, and median ZIP income quartile all had significant impact on the likelihood of a patient being admitted through the ED.

Payer. Medicaid and self-pay admissions had the highest likelihood of originating in the ED. This finding may reflect that outpatient physicians are less willing to care for patients with low or no guarantee of reimbursement, whereby leaving the ED as a more viable care option. In addition, Medicaid patients typically do not have a financial obligation for ED visits and therefore, may opt more quickly for this source of care. The finding that Medicaid and Medicare payer admissions had higher rates of documented SDoH Z-codes is likely a reflection that these benefit offerings are among the first to reimburse or require these data points for value-based arrangements or population health efforts.

Gender. Females characteristically utilize healthcare services at a higher rate than males, possibly due to their proclivity to proactively manage their health^37,38. These results may reflect that females have a lower propensity to be admitted through the ED because they are more engaged in preventative measures, and therefore do not seek care in emergency settings. Females were also less likely to have a documented SDoH Z-code.

Age. Age is significantly positively associated with the likelihood of admission through the ED. This finding is consistent with previous research that has also identified this positive relationship^39,40. Low ED use is common in younger individuals who use fewer health care services in general⁴¹. Interestingly, the mean age for individuals with a documented SDoH Z-code was significantly less than those without documentation. This contradicts comparable research findings²⁸.

Median ZIP Income Quartile.The result that as median income levels rise, the likelihood of having ED admission increase, was not anticipated. However, this is likely representative of those with higher incomes more appropriately leveraging the ED for emergent conditions that warrant triage to the inpatient setting versus patients with lower income levels that are more likely to use the ED for non-urgent reasons^38,41. Further research is needed to explore this finding.

The only clinical covariate included in the analysis was the greatest contributor to the likelihood of ED admission. Due to the complexity of dealing with comorbidities and the higher risk of adverse events, it is expected that individuals with a higher CCI would be more likely to be admitted through the ED. This is further substantiated by previous research⁴¹. In contrast, patients with a lower CCI were more likely to have a documented SDoH Z-code. This contradiction should be explored further.

Social data aggregated at the local, state, and national level could be used for policy decision making and population health management⁴². While previous efforts have shown significant impact to health outcomes and costs, many initiatives could benefit from analyses of these standard social data in aggregate to further inform intervention assessment and comparison^6,43.

Limitations

Limitations of this work include the lack of prior utilization data to control for patients’ utilization history. Furthermore, a single admission may not fully document a patient’s health status. The documentation of Z-codes related to social and economic needs is underutilized and may be more prevalent in sub-populations, making these findings lack generalizability. This study focused on the adult population making associations to pediatric populations limited. Lastly, further research is needed to determine which SDoH Z-codes (or SDoH Z-code families) have the greatest association with admission through the ED. Despite these limitations, this study provides a robust analysis of trends of single ED admissions.

Conclusion

The value shift in healthcare necessitates the understanding of patients’ social, economic, and behavioral needs to supplement clinical information in risk assessment of ED utilization. This study highlights the relationship between hospital admissions originating in the ED and documented SDoH Z-codes using a national sample. The most prevalent SDoH Z-codes pertained to inadequate housing. Adjusting for demographic and clinical patient attributes, the odds of an ED admission increase with the presence of a documented social or economic need. Other significant subgroups that increased the likelihood of ED admission included female patients, older patients, more chronic patients, patients living in more affluent ZIP codes, and Medicaid and self-pay patients. The set of ICD-10-CM SDoH Z-codes that currently serve as a taxonomy for the identification of social and economic needs of patients proved to be valuable in understanding ED utilization. Standard collection of social and economic needs within EHR systems or administrative claims databases can help to target specific resources that meet identified social needs. In addition, social assessments and proper documentation should occur upstream in the primary care setting to identify and alleviate social barriers to care and avoid possible unnecessary ED visits and costs.