The sea of zeros: why cancer patients may be hiding their distress on routine screening forms

Every cancer clinic visit in many hospitals now includes a moment that seems deceptively simple: the patient fills out a tiny questionnaire about emotional distress. One of the most common tools, the distress thermometer, asks patients to rate their distress on a scale from zero to ten, much like a pain scale. Alongside it, many clinics administer the Patient Health Questionnaire-2, a two-item depression screen distilled from the longer PHQ family of instruments. The idea is that ultra-brief screens can catch patients who need psychosocial support without burdening them during busy appointments. But a new analysis of nearly forty thousand screening results suggests that this system may be quietly failing at one of its most important jobs, because far too many patients are answering with a flat zero.

The study, published in the journal Supportive Care in Cancer, was led by Lauren Z. Rynar and Teresa L. Deshields of Rush University Medical Center, with colleagues at Northwestern University, Ohio University, and Roswell Park Comprehensive Cancer Center. The team extracted electronic health record data from outpatient oncology encounters between July 2021 and July 2023, covering 39,762 screening instances in 7,081 patients. What they found was striking: 92 percent of PHQ-2 results were zero, and 91 percent of distress thermometer scores were zero. On its face, that would mean a cancer population almost entirely free of emotional suffering, a claim that contradicts decades of psycho-oncology research suggesting that a large minority of patients experience clinically significant distress at some point in their care.

The statistical heart of the study lies in a class of models designed precisely for situations like this one. When a count-based measure produces more zeros than a standard distribution can explain, researchers turn to zero-inflated models, which split the population conceptually into two groups: one whose true score is genuinely zero, and another whose zero is a spurious or excess zero produced by some other process. The team fitted zero-inflated negative binomial models, which combine a count distribution capable of handling overdispersed scores with an inflation component that estimates how much of the pile-up at zero cannot be accounted for by ordinary variation in distress. The results were dramatic. For the PHQ-2, an estimated 59 percent of the zeros were excess zeros beyond what the underlying distribution would predict. For the distress thermometer, the estimated zero inflation reached a remarkable 90 percent.

Those numbers do not mean that ninety percent of patients lied on their forms. What they mean is that the observed distribution of scores is so dominated by the zero category that standard models cannot reconcile it with plausible variation in true distress. Something in the measurement process, or in the clinical context surrounding it, appears to be pushing scores toward the floor. The authors point to potential underreporting, measurement limitations of ultra-brief instruments used repeatedly over time, and implementation factors such as how often patients are screened and in what setting. When the same two or three questions are asked at every visit, patients may learn to answer quickly and reflexively, and the act of ticking a zero can become a social gesture rather than a considered report.

The models also revealed who was more likely to show genuine distress severity when it appeared. Female patients, Black patients, and those with advanced-stage cancer were all associated with higher estimated distress severity, with p-values below 0.01. These findings align with a broader literature documenting that psychosocial burden in cancer is unevenly distributed, shaped by demographic factors, disease trajectory, and differences in access to support. The fact that the models could detect these gradients at all, buried beneath the avalanche of zeros, underscores the study’s central point: real distress signals exist in the data, but the screening apparatus may be burying them under false negatives.

Perhaps the most intriguing finding concerns time and repetition. The probability of an excess zero score increased both with time since diagnosis and with the number of repeated screens, each with p-values below 0.001. In other words, the longer a patient had been living with cancer and the more times they had filled out the same form, the more likely their zeros were to be statistically suspicious rather than clinically genuine. This pattern fits with what implementation researchers have long suspected about screening fatigue. Routine screening programs, including the standards adopted by the Commission on Cancer that made distress screening a requirement of accredited cancer programs, rest on the assumption that repeated measurement improves detection. This analysis suggests the opposite can happen: repetition may train patients into a default response that the scoring system then mistakes for wellbeing.

There was one factor that pushed in the opposite direction, and it is clinically revealing. Patients who had previously screened positive for depression were about one-third less likely to produce excess zeros on subsequent screens, a highly significant effect. One plausible reading is that once a patient has been flagged, the screening interaction changes: clinicians may ask follow-up questions, patients may take the questions more seriously, or a documented mental health history may make patients more willing to report ongoing symptoms. Whatever the mechanism, the finding suggests that the excess-zero problem is not a fixed property of the instruments themselves but is shaped by the surrounding care context, opening the door to fixes that do not require abandoning routine screening altogether.

The stakes here are not merely statistical. Psychological distress in cancer is not a soft outcome. Large cohort analyses have linked distress to site-specific cancer mortality, and studies of hospitalized patients with advanced cancer show strong relationships between psychological and physical symptoms and health care utilization. Consensus recommendations from the American Psychosocial Oncology Society and the Association of Oncology Social Work have emphasized that screening alone is insufficient, and that it must be paired with triage, referral, and evidence-based treatment. If a substantial fraction of screening results are false zeros, then the entire pipeline from detection to referral may be throttled at its first stage. Patients who need support most may be the ones most likely to slip through with a clean-looking score.

It is worth noting that zero scores are not inherently suspect. Genuine resilience is real, and research on trajectories after adversity, including the influential work of Galatzer-Levy, Huang, and Bonanno on resilience and dysfunction, shows that many people truly do return to low-distress functioning. Previous studies in suicide-risk prediction have also documented excess zeros in count data among primary care populations, indicating that zero inflation is a recognized phenomenon in mental health measurement more broadly. The question is not whether any given zero is false, but whether a screening program can tell the difference. The answer emerging from this dataset is that, at current screening frequencies and with current ultra-brief tools, it largely cannot.

The authors’ conclusions point toward practical reform rather than nihilism. They recommend that programs consider complementary assessment methods and optimized screening intervals to more accurately identify patients who need psychosocial support. In practice, that could mean rotating in longer or more varied instruments, varying the mode of administration, reducing the frequency of ultra-brief screens for stable patients while intensifying assessment around high-risk transitions such as diagnosis, recurrence, or progression to advanced disease, and treating a string of zeros not as proof of wellness but as one data point among many, alongside clinical observation and patient-clinician conversation. The deeper lesson is a humbling one for the era of routine patient-reported outcomes: the number a patient enters into a tablet is not the same as the distress they carry, and a screening system that cannot distinguish a true zero from a false one may be measuring compliance as much as it measures suffering. When zero is not nothing, clinics need better tools for finding out what it actually is.

Subject of Research: Excess zero scores and the validity of repeated ultra-brief psychological distress screening in outpatient cancer patients

Article Title: When zero isn't nothing: excess zero scores in repeated distress screening of cancer patients

Article References: When zero isn't nothing: excess zero scores in repeated distress screening of cancer patients. (n.d.). https://doi.org/10.1007/s00520-026-11326-1

Image Credits: AI Generated

DOI: 10.1007/s00520-026-11326-1

Keywords: distress screening, cancer patients, zero-inflated negative binomial, distress thermometer, PHQ-2, psychological distress, patient-reported outcomes, supportive care, screening fatigue, depression screening, psycho-oncology, electronic health records