Risky Hospitals First: Why Interventions Fail Where Care Systems Are Broken

Every hospital on Earth, from the most lavishly funded academic medical center to the most understaffed district facility, exposes its patients to some possibility of avoidable harm. That uncomfortable truth sits at the heart of a new perspective published in Frontiers in Health Services by Mike English of the University of Oxford and the KEMRI-Wellcome Research Programme in Nairobi, and Charles Vincent of the University of Oxford. The authors argue that health systems quietly absorb—and then normalize—a level of avoidable risk so routine that it becomes invisible to the very people running the system. They give this phenomenon a name: System Tolerated Avoidable Risks in Care, or STARC. And they warn that ignoring it may be one of the biggest reasons why well-intentioned clinical trials, quality-improvement programs and technology rollouts fail, or worse, actively harm the patients they are meant to help.

The argument builds on a framework developed two decades ago by safety researcher René Amalberti and colleagues, including Vincent himself, to describe how all safety-critical systems operate imperfectly. In any hospital, clinicians inevitably deviate in small ways from written procedures. These minor deviations are often seen as enabling efficient work rather than creating risk, and even managers in well-resourced settings treat them as acceptable. The trouble begins when pressure rises. When health workers must complete tasks in limited time, with inadequate resources, while still hitting demanding productivity targets, small deviations become more frequent and gradually harden into accepted workarounds. At first, staff know they are bending the rules and why. Relieve the pressure, and good practice tends to return. But under sustained pressure, workarounds become entrenched, and more substantive deviations emerge that genuinely undermine quality and safety.

Amalberti’s framework describes the initial, slightly degraded state as Borderline Tolerated Conditions of Use, where most deviations go unremarked but an accumulation of small errors can occasionally produce serious harm. Push harder, and a system can slide into what he called the Forbidden Space—a state in which the normal limits of acceptable practice are clearly violated and patients face immediate, direct risk. The COVID-19 pandemic offered a global demonstration of this transition, as surges in severe illness forced forms of care that would normally be judged substandard. But acute shocks are not the only route in. A system lacking sufficient reserves of staff time or materials, what the authors call organizational slack, can drift into this dangerous state through chronic stress alone, such as hospitals where severely ill patients are routinely treated on trolleys in corridors for want of beds.

In high-income settings, the Forbidden Space is understood as an anomaly, something that should never become the default. English and Vincent argue that in many low- and middle-income country hospitals, it effectively is the default. Their evidence comes largely from East African hospital research, including direct observation of 1,848 hours of neonatal nursing in Kenyan hospitals, studies of blood transfusion delays, and ethnographic work on neonatal units. These settings combine dilapidated and overcrowded spaces, broken and unstandardized equipment, reuse of single-use items, medicine shortages, and severe workforce deficits. Interns and newly licensed clinical officers may hold primary responsibility for large numbers of critically ill patients overnight, including performing major surgery. Nurses face extreme patient-to-staff ratios while operating equipment they are unfamiliar with and rationing scarce resources. Support staff may be too few even to maintain basic ward cleaning for infection control.

From a distance, the constant improvisation this forces on staff can look like everyday resilience—the system flexing to keep going. Looked at closely, the authors say, it is better described as an everyday crisis. Health professionals in these settings are trained in best-practice standards, yet the operating conditions routinely prevent them from meeting even their own expectations of good care. Over time, they normalize the workarounds, providing some form of care while often remaining acutely aware that avoidable risks to patients exist. That gap between learned standards and daily reality carries a heavy psychological toll, contributing to burnout, moral distress and emotional disengagement, which in turn makes any behavioral change demanded by an improvement program even harder to achieve.

To capture the spectrum of degradation, the authors sketch five zones of operating conditions, illustrated with workarounds observed in high-mortality Kenyan neonatal units. Zone 1 represents best-practice care with only minor deviations and rare significant harm. Zone 2 involves planned deviations judged to carry minimal risk, such as temporarily reducing vital-signs monitoring for stable patients during workload surges or carefully decontaminating single-use equipment for reuse. Zone 3 brings common, ad hoc workarounds with real cumulative risk: missed measurements of vital signs in severely ill babies, skipped checks on nasogastric tube positioning before feeding, poor hand hygiene, lapses in handover. Zone 4 features deviations in high-risk processes—absent patient identification checks before treatments, missed or delayed diagnostic tests and treatments, inadequate supervision of inexperienced staff. Zone 5 is crisis care, where structured schedules and handovers are abandoned, babies share incubators, invasive procedures proceed without appropriate equipment, and infection control collapses.

The crucial conceptual move is to stop attributing these states to careless individuals. The authors argue it is unfair and pointless to blame staff for cutting corners when the system itself has created the conditions. Paradoxically, keeping some level of service running in a dysfunctional hospital often depends on health workers taking risks that managers or peers in calmer settings would deem unacceptable—so the notion of a forbidden space, they suggest, misses the point, because clinicians frequently have no choice but to manage as best they can. The imbalance between care needs and available workforce capacity, compounded by governance failures, is the driving force. Hence STARC: it is the health system, consciously or not, that creates and tolerates the level of avoidable risk to which patients and staff are exposed.

This reframing has sharp implications for interventional research. If a clinical system’s normal state includes substantial, visible risks to patient safety, then the first imperative is a foundational intervention to improve safety before testing anything new. The authors recommend that any improvement or intervention study in a pressured setting begin with an explicit evaluation of existing quality and safety—in effect, diagnosing which STARC zone the system occupies—followed, in the worst cases, by a pre-intervention period and new resources to stabilize it. They also point to the well-documented infrastructure effect, in which clinical trials inadvertently strengthen local systems by providing resources and motivated research staff, improving outcomes for participants and even non-participants. That transient boost, they argue, shows what sustained system strengthening could achieve, and should be prioritized ahead of or alongside trials, both scientifically and ethically.

Even once a system is deemed stable, the intervention team must ask a second question: what will the intervention itself do to a fragile system? Will a new technology consume staff time needed elsewhere, or be operated by personnel without the competence to use it safely? An intervention may improve the specific aspect of care it targets while quietly increasing risk elsewhere—unintended consequences that intervention teams rarely examine. The authors therefore propose embedding foundational co-interventions that stabilize operating conditions as a designed component of studies in pressured settings, and carrying them into any subsequent scale-up. With a huge and largely unresearched burden of morbidity and mortality from unsafe care in low- and middle-income countries, they conclude that interventions addressing the drivers of STARC may offer a higher return on investment and greater benefit to patients than most other medical interventions. Before asking whether a new treatment works, the field may first need to ask a harder question: is the hospital itself safe enough to find out.

Subject of Research: Patient safety and system tolerated avoidable risks in health care intervention research

Article Title: Is it safe to conduct interventional research? In clinical settings working under pressure we must first consider system tolerated avoidable risks in care

Article References: Is it safe to conduct interventional research? In clinical settings working under pressure we must first consider system tolerated avoidable risks in care. (n.d.). https://doi.org/10.3389/frhs.2026.1950035

Image Credits: AI Generated

DOI: 10.3389/frhs.2026.1950035

Keywords: patient safety, STARC, health systems, low- and middle-income countries, workarounds, quality improvement, interventional research, neonatal care, moral distress, health system resilience, clinical trials, forbidden space