Evaluating Intensity, Complexity, and Potential for Causal Inference in Social Needs Interventions

This review of a scoping review analyzes the intensity and complexity of social needs interventions in randomized clinical trials and whether the trials were designed to assess the effect of individual intervention components on health outcomes.


Introduction
[4][5] Social determinants include both upstream structural and societal systems, policies, and norms and the downstream manifestations of those upstream factors, such as the day-to-day availability of food, transportation, housing, and safety.These downstream manifestations of social adversity are often referred to as social risks. 6More recently, the health care sector has begun to support health care activities focused on reducing social risks (ie, social care or social needs interventions). 69][10] The value of social needs interventions needs to be assessed to understand the extent of health care sector involvement.Although several studies have suggested that social needs interventions undertaken in health care settings may improve health outcomes 11 without increasing (and sometimes even lowering) health care costs, [12][13][14][15] the added value of individual intervention components has received relatively little attention.By design, randomized clinical trials (RCTs) assess the effect of interventions on measured outcomes, but major barriers to uptake of social needs interventions include the lack of resources to implement and sustain these programs. 16,17As a result, assessing value also requires detailed information about the feasibility of implementation to help allocate scarce resources such as staff time, technology, and partnerships.Information about these program costs can be garnered from intervention intensity (eg, duration and extent of patient contacts) and intervention complexity (eg, range of needs addressed, intervention components).
To begin to inform implementation and scalability questions, we undertook this review of a scoping review of RCTs to better understand what the current literature reveals about the intensity and complexity of existing intervention models and the contribution of components of social needs interventions to health and health care utilization outcomes.Although numerous prior systematic and scoping reviews have synthesized the evidence on social care and social needs interventions, [18][19][20][21][22][23][24][25] to our knowledge, this review of a scoping review is the first to focus on these crucial precursors to implementation and scalability.Specifically, we focus on (1) intensity and complexity of social needs interventions and (2) measurement of the effects of individual components or combinations of intervention components on behavioral, health, or health care utilization outcomes.

Data Extraction and Quality Assessment
A reviewer (S.M.K., N.S., V.N., or S.K.) extracted population and intervention characteristics, social needs addressed (eResults in Supplement 1), recruitment and intervention setting, and intervention practitioner; a second reviewer (M.V., M.L.E., or N.S.) checked for accuracy.A reviewer (M.V., N.S., or M.L.E.) assessed the risk of bias using the Risk of Bias 2.0 instrument, 33 and a second reviewer (M.V., N.S., or M.L.E.) spot-checked the studies (eTable 14 in Supplement 1).
To understand intensity of social needs interventions, we extracted information on the number, duration, and frequency of contacts and the time period over which the contacts occurred. 34Given the underlying heterogeneity, we did not define thresholds a priori; instead, we employed an exploratory approach.Specifically, we selected the modal value to categorize the distribution as suggestive of lower vs higher intensity ( <8 contacts vs Ն8 contacts, mode = 8; less often than every 2 weeks vs 2 weeks or more often, mode = 2 weeks; <30 minutes vs Ն30 minutes, mode = 30 minutes; <6 months vs Ն6 months, mode = 6 months).Studies that planned to vary intensity based on participant needs were included in the high-intensity category (ie, varied by need) because they were designed to accommodate high intensity for at least some participants.
To capture features of complexity, we used the Complexity Assessment Tool for Systematic Reviews (iCAT-SR). 35Overall, we applied 5 of the 10 dimensions of the iCAT-SR tool to disaggregate complex interventions (eTable 15 and eTable 16 in Supplement 1).For complexity specifically, we included number of components (iCAT-SR dimension 1), behavior changes targeted in recipients (knowledge, action, or practice) (iCAT-SR dimension 2), and degree of tailoring intended or permitted (iCAT-SR dimension 4). 35Additionally, we assessed whether interventions addressed multiple social needs, had dedicated staff, involved multiple practitioners, and provided resources and/or active assistance with resources or required resources to implement (eg, information, economic supports, food, transportation, supplies, referral for participants, staff, training, time, space, or monetary resources).
We also applied the iCAT-SR framework to assess whether each study's design permitted attribution of effects to 1 or more intervention components.We used 3 specific iCAT-SR criteria, either in combination with study design features or independently.First, we evaluated whether the studies could isolate the effect of the social needs intervention component.Studies comparing usual care plus a single-component social needs intervention with usual care alone permit causal inference on the effects of the single component; factorial trial designs may similarly permit causal inference regarding individual components.Prespecified or post hoc analyses of intervention components may not necessarily support causal inference because they may conflate selection and treatment effects, but they can offer an upper bound on the likely treatment effect. 36

Data Analysis
We relied primarily on descriptive analyses.These analyses were supported by study counts and percentages (from proportions of all studies) and supplemented by qualitative syntheses of individual data elements.

JAMA Network Open | Equity, Diversity, and Inclusion
Evaluating Intensity, Complexity, and Potential for Causal Inference in Social Needs Interventions

Results
We reviewed 15 114 references from database searches, 917 references from the Social Interventions Research and Evaluation Network, and 475 references from hand searches of systematic reviews, for a total of 16 506 references.We excluded 15 010 references at title and abstract review and assessed the full text of 1496 references.We excluded 1419 references at full-text review and included 77 RCTs  reporting on 78 interventions in 93 publications with a total of 135 690 participants (eFigure 2 in Supplement 1).

Intensity and Complexity of Social Needs Interventions
eTable 18 in Supplement 1 characterizes the intensity and complexity of studies along multiple domains.Although reporting of planned intensity was inconsistent, 68 RCTs (88%) reported at least 1 feature suggestive of high intensity (Ն8 contacts, Ն30 minutes per contact, frequency every 2 weeks or more often, duration Ն6 months, or home visits). 36, 40-43, 45-57, 59-69, 71-96, 98, 100-104, 106-112atures of complexity (number of social needs addressed; whether or not a dedicated staff person mediated interactions between patients and the health care system; multiple practitioners, intervention components, behavior targets, resources offered to participants, or resources required to implement the program; and the ability to tailor the program) were more consistently reported than features of intensity.All studies reported at least 1 feature suggestive of complexity and 68 (88%) had 4 or more features that suggested complexity. 36-38, 40-43, 45-63, 65-69, 71-96, 98-106, 108, 112
Planned contacts ranged from a single encounter to encounters at different intervals.46,61,62,77,84,87,89,98,101,102,106,110,112 The number of actual contacts (calls or text messages) ranged from 0 to 681.One intervention evaluating a care coordination intervention planned 24 contacts per participant, but the number of actual contacts ranged from 1 to 90. 41 In 9 studies (12%) that reported on actual but not planned contacts, numbers ranged from no visits to a median of 79 care coordination activities or contacts per client.36,40,62,77,87,89,101,102,106  a Studies may have reported addressing 1 or more of the prespecified social needs that were systematically captured.In addition, studies may have also reported that they addressed any need that arose in the population or social needs that were not prespecified.
b Percentages add up to more than 100% because studies could have included participants in more than 1 age group, recruited from more than 1 setting, or have been conducted in more than 1 setting (eg, the group of studies with only older adults is also included in the group of studies with older adults).
c Defined as more than 50%.
d Studies included at least 1 specified or unspecified chronic condition.
b Percentages add up to more than 100% because practitioner, intervention components, and intervention recipients were not mutually exclusive (eg, more than 1 practitioner could have provided an intervention).
c The effects of the intervention are likely to be transferrable across a limited range of settings only (eg, only within a specific country or health system).
d The effects of the intervention do not appear to be highly dependent on the implementation setting (ie, it is anticipated that the effects of the intervention will be similar across a wide range of contexts or settings).
e Highly, the effects of the intervention were modified by both recipient and practitioner factors; moderately, the effects of the intervention were modified by 1 or more recipient or practitioner factors; independent, the effects of the intervention are not modified substantially by recipient or practitioner factors.
8, 102-105, 110-112 Training encompassed education on motivational interviewing, implicit bias, home assessment, care navigation, financial coaching, and specific conditions such as hypertension.these studies directly address the causal effect of a single social needs intervention component. 81,103,110,111No study reported the use of multiphase optimization strategies. 113garding analyses of intervention components or characteristics, 13 studies (17%) planned or reported subanalyses (a priori or post hoc). 36,37,45,57,59,71,79,86,89,92,96,98,106Specifically, 1 study 57 noted the infeasibility of randomizing all permutations of intervention components and described planned qualitative and quantitative approaches (implementation of components, mediator analyses, and perceptions of staff regarding effectiveness) to assess the effectiveness of components.One study 45  ). 38-44, 46-56, 58, 60-67, 69, 72-78, 80-85, 87, 88, 90, 91, 93-95, 97, 99-103, 105, 107-112As a result, the effects of individual intervention components could not often be distinguished from the potential moderating effects of contextual factors.oaches identified parent strengths and navigated participants to cost-saving services and public benefits 104 Economic supports Subsidized housing 95 Funds loaned to cover apartment security deposit 62 Food Biweekly fresh healthy foods 61 Home-delivered food boxes tailored to nutritional needs and ethnic food preferences every 2 weeks, for 24 weeks 63 Transportation assistance Taxicab vouchers 97 Navigators helped patients access medical transportation assistance through the state Medicaid system 53 Supplies Cell phones 76 Program staff helped clients obtain donated furniture and appliances 62 Referrals to resources or practitioners Accompanied to clinic and introduced to care team 45 Interventionists provided referrals to other health center resources if indicated 59 Other Free primary health care, radiology, and laboratory services 38 Pro bono legal services provided to families with specific legal needs

Measuring the Effects of Individual or Combinations of Intervention
Multicomponent interventions (iCAT-SR dimension 1) and interventions addressing a combination of medical and social needs that have no prespecified or post hoc analyses of the effectiveness of intervention components address overall effectiveness but not the effectiveness of individual social needs components.Second, we judged whether context or setting (iCAT-SR dimension 8) or individual-level recipient or practitioner factors (iCAT-SR dimension 9) were likely to modify the effect of the intervention.Interventions that could be delivered under various settings with minimal modification were assessed as independent of context.Interventions likely to yield different results by setting or fully intertwined within a complex setting were moderately or highly dependent on context.Studies moderately or highly dependent on context and individual factors without additional analyses were judged as being unable to parse the effects of the intervention from the effects of context and individual factors.

Figure 35 36
Figure.Domains Addressed by Social Needs Interventions

Table 1 .
Study and Population Characteristics (continued) Abbreviation: NA, not applicable.

Table 2 .
Intervention Features

Table 3 .
Examples of Standardized Resources and Resources Tailored to Participant Needs 54a Standardized resources may have initially been developed via sociocultural tailoring.Strain, Employment, Transportation, Utilities, Social Isolation, Early Childhood Development, Legal Services, and Childcare (February 1, 2023) eTable 11.Ovid MEDLINE® Search String and Yield for Interpersonal Violence MEDLINE Search (Ovid MEDLINE®) (February 7, 2023) eTable 12. Cochrane Library (Including Both Cochrane Database of Systematic Reviews and Cochrane Central Register of Controlled Trials) Search String and Yield for Interpersonal Violence (February 1, 2023) eTable 13.Inclusion and Exclusion Criteria eMethods.Identification of Social Needs eFigure 1. Screening Approach eResults.Identification of Social Needs eTable 14.Risk of Bias Domains and Ratings eTable 15. iCAT SR Dimensions, Assessment Categories, and Elaboration and Explanations eTable 16.RCT Abstraction Form Items Adapted From iCAT eFigure 2. Article Flow eTable 17.Description of how Social Needs Were Identified eTable 18. Measures of Intervention Intensity and Complexity eTable 19.Justification of Intervention Components eReferences.