Potential Harms and Unintended Consequences of Social Prescribing: What Should Viet Nam Anticipate, Monitor and Mitigate?
EVIDENCE & EVALUATION
Potential Harms and Unintended Consequences of Social Prescribing: What Should Viet Nam Anticipate, Monitor and Mitigate?
Category: Evidence & Evaluation
Social prescribing is usually discussed through a positive logic: understand what matters to a person, connect them with appropriate non-clinical support, strengthen social connection and support health or well-being. That is a reasonable ambition. But an intervention can be well intentioned and still create problems that were never part of its original design.
Someone may be connected to an activity that does not fit their needs. A person who has already struggled to seek help may encounter another disappointment when a community service has no capacity. People facing transport, disability, financial, language or digital barriers may be less able to benefit than those for whom participation is easier. Community organizations may receive more referrals than they can realistically absorb, while navigators may gradually encounter needs beyond the intended scope of their role.
For Viet Nam, this is an important part of developing social prescribing responsibly. The question should not only be “Could this pathway help?” It should also be “How could this pathway unintentionally make things worse, for whom, and how would we know?”
Good intentions do not guarantee harmless implementation
Potential harms received relatively little systematic attention in the social-prescribing literature until 2026. In May, BMJ Open published the first global umbrella review specifically examining potential harms and developing a dark logic model for social prescribing. It included 16 reviews, 14 systematic reviews and two scoping reviews, covering 295 primary research articles. Fifteen reviews reported a combined sample of 106,556 participants. Using a Typology of Harms Framework, the authors considered physical, psychological, equity, group/social and opportunity harms. Opportunity harms were reported most frequently, while the evidence also supported plausible psychological, equity and group/social harms. No direct physical harms were identified in the included reviews.
Those findings should not be interpreted as estimates of how often harms occur. The search ended in June 2024, the umbrella review depended on previous evidence syntheses that were not necessarily designed to identify harm, and the authors explicitly noted that the potential harms identified were unlikely to be exhaustive. The absence of reported physical harm therefore does not demonstrate that physical harms cannot occur.
A second important review, published in July 2026, broadened the picture. Papon and colleagues searched primary studies directly through September 2024. Eighteen primary studies met their inclusion criteria, while a further 87 studies were identified through 35 reviews. Across the material examined, the authors mapped 776 harms, negative or unintended consequences using the CONSEQUENT framework. These consequences extended beyond participants to stakeholders and wider systems, with prominent issues involving health systems, acceptability and adherence, health, and equity.
Together, these reviews do not show that social prescribing is inherently unsafe. They highlight something more useful: the evidence base has historically been much better at looking for benefit than systematically looking for what might go wrong.
Think about the pathway's “dark logic,” not only its intended logic
A conventional logic model asks how an intervention is expected to produce benefit. In social prescribing, the intended sequence might be that a person-centered conversation identifies something important, navigation leads to an appropriate community connection, participation strengthens social connection or meaningful activity, and this contributes to better well-being.
A dark logic model asks what could happen if those same steps do not work as intended. Poorly matched referrals, discontinuity, inaccessible activities, transport or financial barriers, limited community capacity and unclear responsibilities may contribute to disengagement, exclusion, wasted resources or loss of trust. Unintended consequences may therefore emerge from a sequence of smaller implementation failures rather than from one dramatic event.
A person may receive an unsuitable referral, discover that the service cannot accept them, receive little follow-up and eventually conclude that seeking support is not worthwhile. For Viet Nam, this perspective is particularly useful during early implementation. Before asking only how a pathway might succeed, teams should also ask how it could fail, who might be affected and whether those failures would become visible.
Psychological consequences may emerge when pathways repeatedly disappoint people
Current evidence on psychological harm remains limited and should be interpreted cautiously. The 2026 umbrella review identifies plausible pathways rather than reliable estimates of psychological adverse events. Inappropriate referrals, poor continuity and weak alignment between people's needs and the support offered may contribute to disengagement and could plausibly contribute to distress, anxiety, isolation or reduced trust.
The mechanism is understandable. Someone who has already found it difficult to ask for help may be told that a community program could support them, only to discover that it has closed, is unsuitable or has a long waiting list. They may then be referred elsewhere and receive no response. Another person experiencing loneliness may be connected to a group activity that makes them uncomfortable and come away feeling that they have somehow failed at becoming more socially connected.
These examples should not be presented as established frequent harms of social prescribing. They illustrate why mismatched referrals, repeated failed connections, disengagement and negative participant experiences need to be visible in evaluation, rather than disappearing into a generic category such as “did not attend.”
Participation should remain a choice
Person-centered care can become less person-centered if an offer gradually turns into an expectation. A clinician or navigator may genuinely believe that an activity would help. Referral targets can create subtle pressure to convert conversations into referrals, while people themselves may worry that declining an offer will disappoint their healthcare professional or affect their care.
An appropriate outcome of a “what matters” conversation may simply be that the person does not want community support now. They may prefer to manage the issue themselves, reconsider it later or focus on another priority.
A pathway should therefore distinguish declining an offered connection from failing to engage with care. Referral volume should not become a proxy for quality, and a lower referral rate may sometimes reflect more thoughtful person-centered decision-making.
Equity harm can occur even when everyone receives the same offer
Equal referral does not guarantee equal access. The harms literature identifies potential equity consequences arising from variation in workforce skills, service design, intensity of support, supervision and availability of appropriate resources. The broader global evidence also identifies transport barriers, socioeconomic disadvantage, digital exclusion, language and literacy challenges, cultural mismatch, fragmented systems and limited community-sector capacity as recurring implementation problems.
For Viet Nam, these risks can be highly practical. A pathway that relies heavily on smartphones may work well for some urban participants while excluding people with limited digital access. A supposedly free activity may still require transport costs that make participation unrealistic. A person with mobility limitations may receive exactly the same referral as everyone else but have no practical way to enter the venue.
Equity should therefore be evaluated after referral, not only at the point of offer. The relevant question is not simply whether different groups were referred, but whether they could reach, use and benefit from the support under realistic conditions.
Group and social consequences can arise from poor fit and stigma
Community support is shaped by culture, age, gender, language, family expectations and social context. The harms literature identifies potential group or social consequences when services are poorly matched to those realities.
An activity considered appropriate by professionals may not feel socially acceptable to the person being referred. A mixed-gender activity, evening program or service far from home may create barriers that are invisible in a resource directory. The way a program is labelled can also matter. An activity presented specifically for “lonely people,” “vulnerable people” or people with “social problems” may unintentionally reinforce stigma.
This does not mean programs should disguise their purpose. It means communication and co-design should consider how people experience the identity attached to participation, rather than looking only at whether a resource technically exists.
Opportunity harms deserve particular attention in an emerging system
Opportunity harms were the most frequently reported category in the 2026 umbrella review. They include losses associated with costs, inappropriate or ineffective interventions and resources being used in ways that reduce the opportunity to achieve greater benefit elsewhere.
For the individual, participation can require transport, time away from work, caregiving arrangements or other costs. If an activity provides little value, those resources have still been spent. For healthcare, repeated unsuitable referrals consume clinician and navigator time. For community organizations, receiving referrals, responding to queries and providing feedback can create administrative work that competes with their core activities.
At system level, investment in referral technology or large-scale screening can generate demand without creating the community capacity needed to respond. For Viet Nam, where social prescribing remains an emerging field, this is an important lesson: building the referral machinery before understanding whether the pathway works can itself create opportunity cost.
Community organizations can experience unintended consequences too
Recent harms research increasingly emphasizes that unintended consequences do not affect only the person receiving a referral. They may also affect staff, community organizations and wider systems.
A small community organization may begin receiving referrals involving needs far beyond what it was designed to address. Volunteers may feel pressure to respond to complex psychological, financial or family problems. Staff may be asked to complete healthcare-style documentation or return detailed outcome information without additional funding.
As demand grows, a community program could also become more difficult for its existing users to access. This kind of crowding-out effect should be treated as a plausible risk to monitor rather than assumed to be common.
Responsible social prescribing should therefore ask not only whether a community organization exists, but whether it has agreed to participate, understands the referral pathway, has realistic capacity and can receive referrals without undermining its original mission.
Workforce pressures can create risks for participants
Social prescribing depends heavily on relational work, which means workforce conditions can influence participant experience. Inadequate training, unclear role boundaries, limited supervision, excessive caseloads and staff turnover can weaken continuity and create wide variation in the support people receive.
A navigator initially expected to help people access community activities may gradually encounter severe psychological distress, abuse, housing instability, financial crisis or complex family circumstances. Without clear boundaries and escalation arrangements, workers may begin operating beyond their competence or may be unable to respond appropriately when needs exceed their role.
This does not mean navigators should become clinicians or professional social workers. It means organizations need to know what the role can safely do, what it should not do and what happens when a person's needs require another professional response.
Lack of reported physical harm is not proof of physical safety
The 2026 umbrella review did not identify direct physical harms in the reviews it included. These were defined as adverse physical outcomes associated with prescribed activities, such as injury or health deterioration. At the same time, the authors emphasized that harm reporting in the evidence base remains limited and that the findings are unlikely to be exhaustive.
The appropriate conclusion is therefore not that social prescribing has no physical risks. Many pathways connect people with walking, exercise, gardening or other physical activities that can carry some risk depending on the individual and the activity.
For Viet Nam, monitoring should remain proportionate. A book club does not require the same risk management as a strenuous exercise program. The aim is not to turn ordinary community life into clinical care, but to avoid treating silence in published adverse-event data as evidence that relevant risks cannot exist.
Data collection and information sharing can also cause harm
Social-prescribing pathways may collect information about health, disability, family circumstances, finances, housing, loneliness or other highly personal issues. That information may then move between healthcare organizations, navigators and community organizations whose data-protection infrastructure can differ considerably.
Viet Nam's Law No. 91/2025/QH15 on Personal Data Protection has been in force since 1 January 2026. Article 26 requires the data subject's consent for the collection and processing of personal data relating to health, except in the circumstances specified in Article 19(1). Decree No. 356/2025/ND-CP, also effective from 1 January 2026, includes health status and information about private life, personal secrets and family secrets among categories of sensitive personal data.
A community organization rarely needs a person's entire medical or psychosocial history simply because it receives a referral. Information sharing should therefore be purpose-driven and proportionate. Organizations need to consider what information is genuinely necessary, who needs access to it, how it will be protected and what legal requirements apply to the particular processing activity.
Information should follow the function of the pathway, not the curiosity of the evaluator.
Professional boundaries are part of harm prevention
Viet Nam already regulates professional social-work activities. Decree No. 110/2024/ND-CP, effective from 15 October 2024, establishes the broader framework for social work and provides that from 1 January 2027, people practicing social work must hold the applicable social-work practice registration certificate. Circular No. 29/2026/TT-BYT, effective from 25 August 2026, further regulates social-work practice and the updating of professional knowledge.
In hospitals, Circular No. 51/2024/TT-BYT, effective from 1 March 2025, applies to medical examination and treatment establishments licensed to operate in the form of hospitals and regulates social-work services and processes.
These instruments do not establish social prescribing as a profession or national service model in Viet Nam. They do mean that organizations should examine what people actually do, rather than assume that creating a new job title changes the professional nature of the activities being performed.
If a community navigator begins conducting specialized psychosocial assessment, counseling, case management or other activities falling within regulated social-work practice, calling the role a “navigator” does not remove the applicable professional requirements. Clinical assessment and treatment likewise remain within appropriately qualified clinical roles.
Role drift can therefore become both an organizational and a safety problem.
Not every implementation failure is a harm
A mature approach also needs to avoid calling every undesirable outcome a harm. Someone declining a community activity is not automatically harmed. A referral that produces no measurable improvement is not necessarily harmful. Leaving an activity because it no longer fits someone's preferences may simply reflect choice.
It is useful to distinguish lack of benefit, implementation failure and harm, even though the three can overlap. A service may fail to achieve an intended outcome without making anyone worse off. But repeated ineffective referrals can become an opportunity harm when they consume substantial participant time, community capacity or healthcare resources that could have been used more productively.
Likewise, a referral may initially fail because a service has no capacity. That is primarily an implementation problem. If repeated promises of support are then not fulfilled and the person loses trust in seeking help, the consequences may become broader.
Understanding how failure can develop into harm is more useful than labelling every unsuccessful referral as an adverse event.
Harm monitoring should sit beside benefit evaluation
Most social-prescribing dashboards naturally focus on positive measures: referrals, successful connections, participation, satisfaction and well-being. A responsible evaluation should also look for negative signals.
There is not yet a widely accepted and validated harm-monitoring indicator set specific to social prescribing. The two major 2026 harms reviews instead reinforce the need for more systematic identification and reporting of negative and unintended consequences.
An early Vietnamese program can nevertheless monitor proportionately. Potential signals might include inappropriate referrals, repeated unsuccessful connections, complaints or distress following referral, unexpectedly high dropout, accessibility problems, unequal connection rates between groups, community services reaching capacity, excessive waiting times, navigator overload, repeated escalation beyond role boundaries, privacy incidents and physical adverse events where relevant to the activity.
Numbers alone will rarely explain why these events occurred. Short qualitative follow-up with participants, navigators and community partners may help distinguish personal choice from poor service fit, inaccessible design, insufficient capacity or another implementation problem.
The objective is not to prove that every program causes harm. It is to ensure that the pathway is capable of recognizing harm when it occurs.
Learn from people who do not complete the pathway
Evaluation naturally overrepresents people who remain engaged because they are easier to contact, more likely to attend and more likely to complete follow-up measures. People who leave the pathway may provide different information.
Some have simply made an informed choice not to continue and need no further contact. Others may have encountered transport costs, stigma, language barriers, inaccessible design, poor service fit or negative experiences. Where appropriate and proportionate, pilots should therefore learn from non-participation and unsuccessful connections, not only successful ones.
This matters especially for equity. High satisfaction among people who remain in a program can create an overly positive picture if people facing the greatest barriers consistently disappear earlier in the pathway.
Monitor consequences for the community system as well as individuals
A referral pathway can look successful from the healthcare side while placing increasing strain on the system receiving those referrals. Suppose referral volumes increase and connection rates remain high. That may appear encouraging. But if the community organization maintains that performance through unpaid overtime, unstable volunteer labor or reduced access for its existing users, the model may not be sustainable.
Evaluation should therefore consider community workload, referral appropriateness, administrative burden, waiting lists, capacity constraints and whether participation in the healthcare referral pathway is affecting the organization's existing mission.
The receiving community should not become the hidden cost of healthcare's success.
Anticipate harm before launch
Harm monitoring should begin during design rather than only after complaints appear. A Vietnamese pilot could conduct a simple dark logic exercise during co-design. For each major step — identification, offer, referral, navigation, community connection and follow-up — the team could ask three questions: what could go wrong, who could be affected, and how would we know?
What happens if someone cannot afford transport? What happens if the community service is full? What happens when a navigator encounters severe distress? What happens if unnecessary personal information is sent with a referral? What happens if one population consistently has lower connection rates? What happens if community organizations become overwhelmed?
The purpose is not to predict every future event. It is to make plausible failure pathways visible before implementation has become established.
Harm information should lead to decisions
A responsible program should define in advance what kinds of findings require action. One failed referral may need no system change; a repeated pattern of inappropriate referrals probably does. A temporary capacity problem may require adaptation. A serious data breach, significant safety event or sustained pattern of unsafe role expansion may justify pausing part of the pathway while the issue is reviewed.
There is no universal threshold that SPVN should prescribe. Appropriate action depends on the population, setting, severity and likelihood of the problem, and maturity of the program.
The important principle is that harm data should not simply become another dashboard. Depending on what is found, a pathway may continue, adapt, pause selected components or stop.
Discovering during a pilot that something should not be expanded can itself be an important result of responsible implementation.
Safety monitoring should remain proportionate
There is also a risk of going too far in the opposite direction. Social prescribing can become so risk-managed that ordinary community life is unnecessarily medicalized.
A gardening group should not be expected to operate like a hospital ward. A volunteer running a community activity should not become a clinical risk assessor simply because healthcare sometimes refers people to the program. Excessive forms, incident-reporting systems and feedback requirements can themselves create opportunity costs and administrative burden.
The level of governance should therefore reflect the activity, population and reasonably foreseeable risks. What needs to remain consistent is not bureaucracy but clarity: what the activity provides, who is responsible for what, how serious concerns are escalated, what information genuinely needs to move and how unexpected problems are learned from.
What could harm-aware implementation look like in Viet Nam?
An early Vietnamese pilot does not need a complex safety-management infrastructure. It needs a pathway capable of learning from both positive and negative experience.
Before launch, teams can identify plausible unintended consequences alongside intended benefits. Referral criteria and role boundaries should be clear, community partners should confirm what they can realistically provide, and people should understand that accepting community support is their choice. Information-sharing arrangements should be designed before identifiable data begin moving between organizations.
During implementation, teams can review unsuccessful referrals, complaints, unexpected events, inequities in reach, capacity problems and situations requiring escalation. Navigators and community partners should be able to raise concerns without creating a culture in which identifying a problem is treated as evidence that the program has failed.
At regular intervals, the team can ask whether the balance still makes sense. Are people receiving meaningful support? Are particular groups being missed? Are community organizations becoming overloaded? Are navigators moving beyond their intended scope? Is every piece of information collected still necessary? Are resources being spent on components that add little value?
In this way, harm monitoring becomes part of iterative learning, rather than a compliance exercise added after the model has already been established.
Do not measure only what social prescribing improves
Social prescribing has an appealing narrative. It connects healthcare with community life, shifts attention toward what matters to people and recognizes that health is shaped by much more than medical treatment.
Those strengths should not make the field reluctant to examine unintended consequences.
The emerging evidence does not show that social prescribing is inherently unsafe. It also does not justify assuming that the absence of documented harm means a pathway is inherently safe. The more balanced conclusion is that social-prescribing interventions can generate intended and unintended consequences at the level of individuals, workers, organizations and systems, while the evidence about harms remains incomplete.
For Viet Nam, this creates an opportunity to build harm awareness into implementation from the beginning rather than adding it after scale-up.
The mature question is therefore not simply:
“Does social prescribing work?”
It is:
“For whom does this pathway help, for whom might it fail or create unintended consequences, what does it require from the people and communities around it, and can we recognize problems early enough to respond?”
If social prescribing in Viet Nam can ask both sides of that question, benefit and harm, it will be better positioned to develop as genuinely responsible community-connected care.
References
Cooper M, Okeowo D, Bennett L, et al. Potential harms of social prescribing: a global umbrella review and dark logic model. BMJ Open. 2026;16(5):e108998. doi:10.1136/bmjopen-2025-108998.
Papon V, Schöpf J, Litt JS, et al. Harms and Negative or Unintended Consequences of Social Prescribing: A Scoping Review. Healthcare. 2026;14(13):1947. doi:10.3390/healthcare14131947.
Kisa A, Kisa S. Models, implementation, and reported outcomes of social prescribing interventions: a scoping review. Frontiers in Public Health. 2026;14:1871347. doi:10.3389/fpubh.2026.1871347.
World Health Organization Regional Office for the Western Pacific. A toolkit on how to implement social prescribing. Manila: WHO Regional Office for the Western Pacific; 2022.
Government of Viet Nam. Decree No. 110/2024/ND-CP on Social Work. Effective 15 October 2024.
Ministry of Health of Viet Nam. Circular No. 51/2024/TT-BYT on the implementation of social work in medical examination and treatment establishments. Effective 1 March 2025.
Ministry of Health of Viet Nam. Circular No. 29/2026/TT-BYT on social-work practice and updating social-work knowledge. Effective 25 August 2026.
National Assembly of Viet Nam. Law No. 91/2025/QH15 on Personal Data Protection. Effective 1 January 2026.
Government of Viet Nam. Decree No. 356/2025/ND-CP detailing certain articles and measures for implementation of the Law on Personal Data Protection. Effective 1 January 2026.