Background Unmet basic resource needs, such as difficulty affording healthcare, medications, food and housing, may contribute to worse healthcare quality indicators, but interventions are hampered by lack of specific knowledge regarding the distribution of unmet basic resource needs and their association with priority clinical conditions and health service use patterns.
Methods Cross-sectional study of primary care patients in two urban academic practices from 1 October 2013 to 30 April 2014. Patients were screened for unmet needs and enrolled in a programme to link them with community resources. Key measures included patient report of unmet basic resource needs, clinical conditions prioritised by quality improvement programmes (hypertension, diabetes and depression), and health service use patterns such as frequent emergency department (ED) visits (>2 in the preceding year) and frequent clinic ‘no-shows’ (>1 in the preceding year).
Results 416 patients with unmet needs were included, and compared with 2750 patients who did not report needs. The most common types of needs reported were: difficulties affording healthcare (46.5%), food (40.1%) and utilities (36.3%). Patients who reported unmet needs were more likely to have depression (17.8% vs 9.5%, p<0.0001), diabetes (32.7% vs 20.4%, p<0.0001), hypertension (54.3% vs 46.3%, p=0.002), be frequent ED users (11.3% vs 5.4%, p<0.0001), and have frequent ‘no-shows’ to clinic (21.6% vs 11.9%, p<0.0001).
Conclusions Difficulty affording healthcare and food are particularly common needs among patients with priority conditions. Strategies to identify and address unmet needs as part of routine care may be an important way to improve healthcare quality.
- Chronic disease management
- Healthcare quality improvement
- Patient-centred care
Statistics from Altmetric.com
There is widespread consensus that primary care delivery needs effective ways to increase quality and become more patient-centred.1 ,2 Much quality improvement work focuses on factors within the healthcare system, such as processes of care.2 However, there is increasing recognition that the social and economic context in which patients live, get sick and manage their illnesses3 is a key health determinant. This has led to calls for addressing social determinants of health as part of quality improvement, but implementing this has been challenging.4–7
A subset of social determinants of health that may be particularly relevant for quality improvement programmes is unmet basic resource needs—such as difficulty affording food, medications or housing. Understanding how these needs are distributed among patients with clinical conditions and health service use patterns emphasised by quality improvement programmes, such as diabetes, hypertension, depression and frequent emergency department (ED) use, is vital for programme development, but little specific information currently exists.
Further, existing programmes that address unmet basic needs often take a complex-care management approach. This approach uses a comprehensive initial assessment, usually conducted by case managers or community health workers, to ascertain basic resource needs, social support and care coordination issues8–10 and craft an individualised care plan. While often effective for improving health,9 ,10 the expense of these programmes may limit their use to only the most complex patients and the highest users of health services.9
Less intensive and costly programmes are needed to address unmet basic resource needs in a broader range of patients. These unmet needs are key determinants of the health of populations, and likely contributors to recognised disparities in health outcomes among racial/ethnic minority groups, those of lower socioeconomic status and those with limited English proficiency.11–14 Addressing unmet basic resource needs may be an important way to improve care quality, but many healthcare systems do not yet have the skills and knowledge to do this. Relying on clinicians to identify patients with unmet basic resource needs and then refer them to social workers or resource specialists may under-identify eligible patients, overburden clinicians and runs counter to population health principles of proactive patient identification and outreach. Strategies to address unmet basic resource needs as part of routine care should efficiently identify needs that are relevant to quality improvement programmes and enable healthcare providers to practice more effectively.
As part of an effort to identify and address the needs of primary care patients, we examined the association of unmet basic resource needs with medical conditions and patterns of care usage. The goal was to inform the further development of such programmes by determining where there may be ‘room for improvement’ by addressing basic needs, and to identify specific needs for quality improvement programmes to emphasise.
Setting and study participants
Data for this cross-sectional study were collected from two academic, hospital-based primary care practices within a practice-based research network in Boston, Massachusetts. These clinics had an existing complex-care management programme for high-risk Medicare patients.15 All patients empanelled with resident physicians in the practices16 and seen in clinic from 1 October 2013 to 30 April 2014 were eligible for inclusion. There were no exclusion criteria.
This study received institutional review board approval for the use of electronic data captured in the course of routine care without informed consent.
Identifying basic resource needs
This programme was set in two primary care clinics as a way to provide education on and assistance in addressing unmet needs for patients. The programme was administered by Health Leads (https://healthleadsusa.org/), a community-based healthcare organisation that helps connect patients with local services to meet basic resource needs. During routine clinic visits, patients received a form to complete at check-in (see online supplementary digital content 1—Screening Form), which assessed needs in the following domains: health (including difficulties with obtaining health insurance and affording medications), employment (including job training and placement), financial (such as Supplemental Security Income or emergency cash grants), food (such as Supplemental Nutrition Assistance Program (SNAP), formerly the Food Stamp programme, and food pantries), housing (including affordable housing and emergency shelter), legal (including benefits denial and pro bono representation), transportation (both medical and social) and utilities (including subsidies and shut-off protection).
Patients who indicated that they would like help with any of the listed needs had this information given to their physician, who referred them to a service connection programme ‘advocate’ to discuss their needs more fully. Physicians could also refer patients to the programme using email or messaging through the medical record if they recognised a need outside of the screening process.
Advocates staffed a ‘desk’ located in the clinic, and patients could work with the advocate in person, or arrange telephone or text message follow-up. All clinic staff were oriented to the methods and goals of the programme at the beginning of the study period, and encouraged to refer patients to it.
Programme development included conscious efforts to enhance accessibility for those with language barriers, low literacy and cognitive impairment. Screening instruments were available in English, Spanish, Haitian Creole and Chinese, which includes >98% of patients seen in the study practices. To promote access to the screening tool for patients with low literacy levels, the tool used icons, simple sentence structure and repetition of key words. For participants with known cognitive impairment necessitating guardianship, the guardian was the point of contact. Participants with cognitive impairments (such as developmental, language based, executive functioning, acquired brain injuries or neurodegenerative illnesses) who required adaptation of programme methods but were not under guardianship were ‘flagged’ by the programme manager and assigned to more experienced programme personnel (usually a masters in social work intern).
Clinical characteristics and healthcare usage
The goal of this study was to determine the specific association between unmet resource needs and clinical characteristics and healthcare use patterns prioritised by quality improvement programmes,2 including depression, coronary heart disease, cerebrovascular disease, hypertension, diabetes and chronic kidney disease. The presence of these conditions was determined using either previously validated algorithms16–18 or newly validated algorithms (see online supplementary digital content 3—Algorithm Validation). We assessed chronic disease management by determining whether the most recent low-density lipoprotein cholesterol level was below 100 mg/dL for patients with coronary heart disease or diabetes,19 and whether the most recent haemoglobin A1c (HbA1c) was below the commonly used thresholds of 9.0% and 8.0% for patients with diabetes19 using data from an electronic health data repository.20 We also assessed whether patients had received recommended clinical preventive services such as colorectal cancer screening (among patients aged 52–75 years without history of prior total colectomy), breast cancer screening (among women aged 42–74 years without history of bilateral mastectomy) and cervical cancer screening (among women aged 21–64 years with no history of total hysterectomy).21 Indicators of adverse usage included high ED use (>2 visits in calendar year 2013), and high number of ‘no-shows’ to clinic visits (>1 in calendar year 2013).
Addressing basic resource needs
In addition to the primary objective of determining the distributions and associations of basic resource needs, we also wanted to determine the ability to ‘link’ patients to resources to meet those needs. To do this, we created a framework to determine whether participants were successfully matched to community resources for their needs. Patients who enrolled in the programme worked with their advocate to define specific areas requiring assistance. These were categorised into 42 ‘sub-needs’ within 8 need categories (see online supplementary digital content 2—Classification of needs and definitions of resolution). The advocate then worked to link the patient with resources to meet the needs, considering factors such as eligibility, desirability and accessibility. For example, a patient reporting a food need who was interested in applying for SNAP would complete an eligibility check with their advocate, be informed of the required verification documentation and be given assistance by their advocate throughout the application process. The advocate would contact the patient weekly until the ‘SNAP’ food sub-need was considered resolved, the patient indicated he or she no longer wanted assistance meeting this need, or the patient was lost to contact. In order for the SNAP sub-need to be considered resolved, the advocate would confirm that the patient was in possession of an electronic benefit transfer (EBT) card with a balance of the appropriate benefit amount. Needs could be closed as successful if they met a priori success criteria, equipped, if the patient indicted that he or she had all the information needed and did not want further assistance, failure, if success criteria were not met, or disconnected, if the patient was lost to follow-up.
The patient's primary care physician was provided with an intake narrative of the patient's identified needs and psychosocial factors related to their situation. Data on distribution, follow-up and resolution of needs were collected and managed by Health Leads.
We considered covariates that may confound the association between unmet needs and clinical characteristics and healthcare use patterns. These included age, gender, self-reported race/ethnicity, educational attainment (dichotomised as<high school diploma vs high school diploma or greater), health insurance (commercial, Medicare, Medicaid, none/self-pay), primary language (English/non-English) and Charlson comorbidity score.
Patients who screened positive for unmet resource needs were compared with all patients who were seen in the clinic during the study period, but who did not indicate an unmet need (either because they screened negative or because they were not asked). Because this group likely contains some patients with unmet needs who were not screened, the analyses we present represent conservative estimates of associations.
We first performed descriptive statistics, and determined which needs were most common among those reporting any needs. We next sought to determine which clinical conditions and usage patterns were more common in patients with unmet needs, which can suggest where there may be ‘room for improvement’ in the management of these situations by addressing basic needs in routine care, and inform areas where assistance programmes may want to specialise. For unadjusted comparisons, we used χ2 tests for dichotomous variables and Wilcoxon tests (given non-normal distributions) for continuous variables. Given that our purpose for these analyses was to answer the programmatic question of which clinical situations had patients with greater needs in order to plan programming to address the need, we did not conduct adjusted analyses for this question. To follow up this question, however, we sought to determine which specific needs were more common in particular clinical and health service use situations, accounting for the sociodemographics and comorbidity of the participants. To help answer this question, we used multivariable logistic regression. By examining a large number of needs that may be associated with a particular clinical and health service use situation, these analyses help provide guidance as to which ‘resource landscapes’ are most salient for particular clinical and health service use situations. This can be useful for programme design when seeking to address these specific situations. All analyses were conducted using SAS V.9.3 (SAS Institute, Cary, North Carolina, USA).
From 1 October 2013 to 30 April 2014, 416 patients with unmet needs were identified and enrolled in the Health Leads programme. Compared with 2750 patients who did not report unmet needs, patients with unmet needs were more likely to be women, racial/ethnic minorities and have Medicaid insurance (table 1). The most common types of needs reported were: difficulties affording or receiving healthcare (including medications), difficulties affording food and difficulties affording utilities such as heating, electricity and telecommunications services (table 2).
The prevalence of depression, diabetes and hypertension was greater in patients with unmet needs, compared with those who did not report them (table 3). Additionally, patients who reported unmet needs were more likely to be frequent ED users, have frequent ‘no-shows’ to clinic appointments, and have poor glycaemic and cholesterol control. The prevalence of being up to date on cervical, breast or colorectal cancer screening did not vary by need status.
Looking at specific needs adjusted for age, gender, race/ethnicity, insurance, educational attainment, primary language and Charlson score (table 4) revealed heterogeneity of types of needs across clinical conditions and usage patterns. Depression was associated with difficulties affording and accessing healthcare (adjusted OR (aOR)1.65, 95% CI 1.07 to 2.54, compared with patients without depression), and difficulties with employment (aOR 2.19, 95% CI 1.18 to 4.04). Diabetes (aOR 1.70, 95% CI 1.15 to 2.51) out of control glycaemia (aOR 1.93, 95% CI 1.19 to 3.15), and hypertension (aOR 1.83 95% CI 1.20 to 2.80) were associated with food needs. Frequent ‘no-shows’ to clinic were associated with housing (aOR 2.15, 95% CI 1.23 to 3.75), transportation (aOR 1.96, 95% CI 1.06 to 3.63) and legal needs (aOR 3.34, 95% CI 1.28 to 8.68).
With regard to programme performance, the mean number of needs per enrolled patient was 2.6. Overall, 62% of needs were closed as successful or equipped, 4% were closed as failure and 34% were closed as disconnected. Resource connections were most frequent for health-related needs (most commonly prescription assistance (35% of cases) and adult health insurance (15%)), food (most commonly pantries and soup kitchens (56%) and SNAP (30%)), and utilities (most commonly electric, gas and oil discount rates (49%) and energy assistance/subsidies (27%)). These were also the most common needs that could not be resolved. Cases were open for a median of 48 days (IR 31–80 days). The median number of participant contacts (a dialogue with a participant, not simply leaving a message) was six.
Among primary care patients seen in an urban academic setting, we found that affording healthcare and food were particularly common unmet needs. Of several clinical conditions and usage patterns, individuals with diabetes, depression and frequent ED use were more likely to have unmet needs. The distribution of particular needs and conditions or usage patterns often suggested rational targets for interventions—for example, food needs were common in patients with diabetes, and transportation needs were common in patients with frequent ‘no shows’. We did not observe important differences in the ability to link patients to community resources for particular needs; that is, there was no particular need for which it was ‘easier’ or ‘harder’ to achieve linkage.
This study is consistent with and adds to the literature demonstrating clinically relevant associations between unmet needs and clinical conditions,22–27 particularly by adding specificity to our understanding of what needs patients face in common clinical situations. Pilot models of universal screening of basic resource needs have been implemented in paediatric settings,28–30 but we are unaware of prior studies in the adult primary care setting. Existing complex-care management, community health worker or patient navigation programmes often assess for the basic resource needs of patients, but efforts to address them are often not differentiated from other aspects of the programme, such as providing education, care coordination and social support.10 ,31–39 Thus, despite growing experience in providing interventions for vulnerable patients, there are few data for healthcare systems to reference when planning quality improvement interventions to help patients meet basic resource needs. Additionally, by providing a more comprehensive assessment of the problems patients face, rather than focusing only on a specific need, such as food insecurity alone,40 and by evaluating a number of clinical situations, rather than a disease-specific management programme,38 ,41 this study provides information about overlapping and competing demands across a range of diagnoses and care usage patterns, and adds more precise data about the associations of particular needs and clinical situations. It also provides another delivery model for integrating social determinants of health information into clinical care.42
Challenges in implementing this intervention included competing demands for clinic staff time, diversity of participant needs and abilities, and co-ordinating with other ongoing efforts to improve patient care. We believe strong leadership, efforts to engage all staff members and clinical providers, incorporation of feedback from both participants and staff, and a shared sense of the importance of addressing basic resources needs contributed to intervention implementation. Work examining how best to implement programmes that address unmet basic needs, using the model described here or other models, is certainly warranted as clinical care moves towards a more holistic approach.
Beyond specific interventions and programmes within clinical settings, these findings have important implications for health policy and for future research. High rates of difficulty affording medication and healthcare, despite near universal healthcare coverage in Massachusetts, suggest that ‘under insurance’ may remain a significant barrier for many chronically ill patients even after implementation of the Affordable Care Act (ACA), especially given the growth in high-deductible health plans,43 ,44 that are often not optimal for patients with chronic illness. As a result, decreasing funding for safety-net programmes after ACA implementation may be premature,45 ,46 and in fact may become more important as newly insured patients seek care outside of traditional safety-net settings. For example, enrolment in a state-sponsored plan called the Health Safety Net,47 which provides benefits for uninsured and underinsured residents, was responsible for the majority of the successful prescription medication and healthcare coverage assistance in this study.
Health information technology may help expand routine basic needs screening within healthcare systems. Visit-independent, universal patient entered report of unmet needs could be solicited, perhaps using an electronic patient ‘portal’, available through mobile technology to give greater access to disadvantaged patients.48 Screening a population to identify those patients at high risk for unmet needs and then directing ‘higher-touch’ targeted outreach represents a type of tiered intervention strategy that has been successfully used in other population management programmes.21 In fact, our finding that basic resource needs were not clinically or statistically significantly associated with completion of cancer screening in this population may be due to the success of this approach in reducing socioeconomic disparities in cancer screening in our primary care network.49
This study should be interpreted in the light of several important limitations. It is too early to know if this programme's efforts to address unmet resource needs led to significant clinical improvement. Thus, these data cannot be taken as evidence that the strategy of helping patients meet basic needs as part of routine care improves health. However, these data are still useful to inform interventions, and point to situations where such a strategy may, or may not, lead to significant changes. Regardless of clinical outcomes, addressing these needs may improve patient well-being and satisfaction with care.
The relatively small number of data points available for analyses of associations between particular needs and clinical situations is another limitation. CIs were often wide, and could not exclude clinically important associations in some situations. A third limitation is that negative screenings from assessments were not consistently collected early in the implementation of the programme, and some eligible patients may have gone unscreened. Thus, our comparison group of patients seen in the clinic during the same time period included both patients who truly did not have unmet needs and patients who did not report needs because they were not asked. Because this group likely contains some patients who would have reported needs had they been screened, this would tend to bias our results towards the null. Next, we did not have access to data on inpatient length of stay or institutionalisation, which would be relevant health service use outcomes to evaluate in future studies. Finally, despite efforts to make the programme accessible to all participants, those with lower literacy levels, cognitive impairment or higher burdens of needs may have been missed in screening due to difficulty using the screening instrument or stigma related to reporting needs. This would tend to bias the observed associations towards the null. Tools to assess literacy and cognitive impairment could be employed in subsequent programmes in order to identify those for whom the standard methods of needs assessment may be suboptimal.
These limitations are balanced by several strengths. Needs were ascertained using practical, real-world screening rather than research definitions, which minimises respondent burden and identifies patients who want services. Additionally, patients were screened for a wide variety of possible unmet needs, and screening was not restricted to a particular diagnosis or criteria such as prior high medical usage. Finally, the non-safety-net setting means that the findings are likely generalisable to numerous American primary care clinics.
Unmet basic resource needs are common among primary care patients, particularly for those with depression, hypertension, diabetes, frequent ED visits and clinic ‘no-shows’. Clinicians often have few mechanisms to address these needs in their patients, and existing models to assist with this, such as complex care management, not be feasible for most primary care patients. These unmet needs are important to patients and are recognised clinical priorities for quality improvement and efficient resource use. Therefore, adding strategies to identify and address unmet needs in all, not only the most complex, patients may be an important way to expand our ability to improve care quality.
The authors would like to acknowledge the contribution of Steven McDermott, RN, for assistance with validation of the electronic algorithms, and Mark Marino, BA, for administrative assistance in operations for the Health Leads programme during the time of the study.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
- Data supplement 1 - Online supplement
Competing interests Three authors (ACH, BJS and KJT) are employees of Health Leads.
Ethics approval Partners Healthcare Human Research Committee.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.