Socio-demographic factors, tumor characteristics, treatments administered, and survival in cancer patients in a prison hospitalization unit: a retrospective cohort comparison with matched community controls.

Introduction
Cancer is one of the main causes of morbidity and mortality in the prison population worldwide [1, 2]. In prisons in the United States, for example, up to 30.2% of disease-related deaths are attributed to malignant neoplasms, surpassed only by cardiovascular diseases in certain settings [3, 4]. Despite the constant health surveillance in the penitentiary context, which could theoretically favor earlier diagnosis, this potential is often not realized [5]. In this setting, several studies indicate that this theoretical advantage does not translate into effective cancer screening or earlier diagnoses, leading to a similar or even higher proportion of advanced-stage cases at the time of initial diagnosis [6, 7].
Moreover, the incarceration of people with serious illnesses raises important concerns about whether prisons can provide appropriate medical care. In many cases, it is difficult to deliver timely and effective treatments, and health conditions often get worse in penitentiary patients. For that cause, early medical care and prevention in the community are so important. This situation also raises deeper ethical questions about justice, fairness, and how we treat vulnerable people. In response, abolition medicine calls for more than reform, it pushes for healthcare systems that move away from punishment and instead focus on dignity, community care, and repairing harm [8–11].
In Spain, prison healthcare is legally required to match the standard of care in the general population. However, in practice, cancer care in prisons often faces delays in diagnosis, limited access to specialists, and interruptions in treatment. Factors like restricted mobility, lack of privacy, and dependence on non-medical staff for symptom reporting can affect timely care. Barriers to early detection and access to oncological treatment in prisons include: logistical and security limitations (transport, restrictive protocols), limited implementation of screening programs (mammography, Papanicolaou tests, fecal occult blood tests, and colonoscopies), a higher prevalence of risk factors (smoking, substance use, coinfections such as HIV or hepatitis), and low socioeconomic and educational levels, which often result in limited prior access to healthcare [12–14].
In contrast, cancer patients in the community benefit from faster referrals, coordinated multidisciplinary teams, and stronger support systems. In regions like Catalunya, efforts such as integrating prison health with the public system and using liaison nurses help improve continuity of care after release [15].
In Spain, no studies have been conducted on cancer incidence among incarcerated individuals, highlighting the need to understand the epidemiological profile, stage at diagnosis, and survival in this population to design targeted screening interventions and improve continuity of care. This study was conducted in a hospital with a unique care model, with a prison hospitalization unit (PHU) physically connected to a community hospital, facilitating comprehensive and continuous medical attention for patients who are incarcerated.
Therefore, this study aimed to analyze the sociodemographic characteristics, tumor types, treatments administered, and survival of cancer patients in a PHU, comparing the diagnostic-therapeutic management of the most prevalent tumors in patients who are incarcerated with patients from the general population.

Materials and methods

Study design and participants
This is a retrospective matched cohort study comparing cancer patients in a Prison Hospitalization Unit (PHU) with community-based hospitalized patients treated in a community hospital as a control group, using electronical hospital databases. The inclusion period covered 15 years (2008–2023), involving individuals who are incarcerated from various prisons referred to our center (PHU) as their reference hospital, all with a confirmed diagnosis of cancer.
PHU at Hospital Universitari Consorci Sanitari de Terrassa (CST) has operated since 1992 as the official inpatient referral center for the penitentiary healthcare system of Catalonia. Integrated within the general hospital, the PHU consists of 33 beds specifically allocated for patients who are incarcerated from five correctional facilities in the Barcelona area: Brians 1, Brians 2, Quatre Camins, the Youth Prison, and the Women’s Prison.
When an incarcerated individual presents with urgent medical needs or requires specialized diagnostic testing, they are first assessed at their designated local hospital. If hospital admission becomes necessary, all referrals from the listed prisons are centralized at the CST PHU. This model promotes standardized inpatient care and enhances coordination across the prison healthcare network, and it is unique in Spain.
CST maintains continuous, structured communication with penitentiary institutions through a specialized and confidential clinical information system, which manages the medical records of patients who are incarcerated. Additionally, CST healthcare professionals conduct regular on-site follow-ups at the correctional facilities when needed. This integrated care model ensures continuity of care, equity in healthcare delivery, and timely access to specialized services for incarcerated individuals.
All oncological therapies for patients who are incarcerated—such as chemotherapy, biological therapies, radiotherapy, and palliative care—were administered at our hospital facilities, under the supervision of the Oncology Department or Internal Medicine Department, in coordination with the PHU staff. The PHU is managed by internists and nursing staff, and patients benefit from integrated care with the hospital’s specialist services. Coordination is facilitated by a PHU manager who oversees diagnostics, referrals, and interdepartmental communication.
From the initial 131 cancer cases identified in the PHU during the study period, the sample was selected to include only the most prevalent tumor types. A total of 78 patients with head and neck, lung, lymphoma, gastrointestinal (which included gastric and small bowel tumors), prostate, and testicular tumors were initially selected. For each of these cases, two individuals were randomly selected from the hospital’s electronic database to form the community-based control group (n = 156), matched by age (± 5 years) and tumor histological type. Patients with presence of synchronous primary tumors at diagnosis were excluded. Additionally, all lymphoma cases were excluded due to incomplete data in the hospital’s electronic records. This included three patients without follow-up data, one individual released from prison with no documented clinical continuity, one patient treated exclusively at another center, and one case lacking sufficient information on treatment, follow-up, and discharge. These cases were excluded to ensure data reliability and consistency in the comparative analysis. Regarding the total of PHU population (n = 131) we have obtained, from major to minor percentage, gastrointestinal cancer (12%), followed by prostate cancer (9%), head and neck cancer (9%), lymphoma and testicular cancer (both 8%), and finally lung cancer (6%). Other, less frequently observed tumor types included colorectal cancer (5%), kidney (3%), breast (3%), gynecological (2%), and, with only 1% each, leukemia, liver, brain, and thyroid cancers.
The final PHU cohort consisted of 63 individuals who are incarcerated with cancer. After excluding 11 lymphoma and 4 synchronous cases from the PHU group, and 22 lymphoma and 5 synchronous cases from the community group, the matching process was re-run within each histological stratum to maintain the planned 1:2 proportion. For every eligible PHU case (n = 63), two community-based controls were randomly selected from the remaining pool, matched by tumor type and age (± 5 years). This procedure resulted in a final distribution by tumor type: PHU (GI 16, prostate 13, head and neck 13, testicular 12, lung 9) and controls (GI 34, prostate 26, head and neck 27, testicular 22, lung 20), totaling 63 and 129 respectively.
Matching was based on age (± 5 years) and tumor histological type. Tumor grade and stage were analyzed separately not being used as matching variables (Fig. 1).

The unit of analysis was the individual patient. Data extraction was performed through manual chart review of clinical notes, pathology reports, and discharge summaries, complemented by queries of structured fields within the hospital’s electronic medical record system. Data abstraction followed a standardized protocol with predefined variables to ensure consistency across reviewers.

Variables
The variables analyzed in this study included information on sociodemographic characteristics (age, sex, nationality); risk factors (smoking, alcohol consumption, and the use of other drugs, such as opioids and cocaine) measured as a dichotomous yes/no response. Disease stage and tumor characteristics were defined by tumor site (lung, gastrointestinal, head and neck, prostate, and others), histological type, tumor size, lymph node involvement, and presence of metastases. Tumor size at diagnosis (in centimeters) was compared between PHU and community-based patients. Lymph node involvement and metastatic disease were assessed as dichotomous variables (yes/no), based on diagnostic imaging and pathology reports.
Given the heterogeneity of tumor types and the limitations of our sample size, we classified tumor stage into three clinically meaningful categories based on components of the Tumor Nodes, Metastases (TNM) classification, where T (Tumor) Indicates the size and extent of the primary tumor, N (Nodes) Describes the spread to regional lymph nodes and M (Metastasis) Refers to the presence of distant metastases:

Early stage: tumors without nodal involment (N0) or distant metastasis (M0). Commontly corresponding to stage I and some stage II. Most often treated with curative intent.

Locally advanced stage: tumors with nodal involment (N+) but no distant metastases (M0). Generally corresponding to some late stage II and stage III. Most often treated with curative intent.

Metastatic stage: tumors with distant metastases (M1). Most often stage IV and treated with palliative intent.

We conducted a dichotomized comparison between early and locally advanced stages versus metastatic stage, in line with prior literature supporting this approach in mixed cancer populations [16, 17].
Regarding oncological treatments, data were extracted from the hospital’s electronic medical record system, including oncology treatment plans, surgical reports, and radiotherapy and palliative care prescriptions. Treatments were recorded as dichotomous variables (yes/no) for each modality.
Routes to cancer diagnosis were also documented based on clinical records and included the following categories: incidental diagnosis, presence of constitutional syndrome at diagnosis, or admission due to warning signs of a specific neoplasm.
Clinical outcomes were assessed through prospective follow-up during hospitalization and subsequent manual review of electronic charts after discharge (direct follow-up). Deaths were categorized as in-hospital or post-discharge mortality. For incarcerated patients, post-release mortality data were obtained from the hospital registry and, when available, validated through linkage with regional health databases (Catalan Health Service mortality records). This dual approach minimized the risk of underreporting. Moreover, since part of the medical team from our hospital also provides clinical care within the prison facilities of Catalonia, this facilitates continuity of care, enhances survival tracking, and contributes to centralized documentation within the hospital’s medical records. Nonetheless, loss to follow-up after release remains a known limitation in this population.

Statistical analysis
Descriptive statistics are presented as absolute and relative frequencies for qualitative variables, whereas quantitative variables are expressed as the means and standard deviations (or medians and interquartile ranges, as appropriate). For bivariate analysis, the chi-square test or Fisher’s exact test was used; for the comparison of means, Student’s t test or nonparametric tests were used if the normality assumption was not met.
We compared sociodemographic factors (age, sex, nationality, smoking, alcohol and other substance use), tumor characteristics (tumor size, lymph node involvement, metastatic disease), stage at diagnosis (early/locally advanced vs. metastatic), and treatment modalities (surgery, chemotherapy, radiotherapy, biologic therapy) between PHU and community-based patients. Overall survival and cancer-specific mortality were also compared between groups.
Kaplan‒Meier survival curves were constructed and compared with the log-rank test (p < 0.05 was considered statistically significant). Cancer-specific mortality was assessed as the number and proportion of deaths within 10 years of diagnosis, comparing PHU patients and matched controls using Kaplan–Meier survival analysis. The data were processed via SPSS v29 software (IBM Corp., Armonk, NY).

Ethical considerations
This study adhered to bioethical standards for research on human subjects set forth in the Declaration of Helsinki. The data were handled confidentially through coded identifiers, and the study protocol was approved by the Research Ethics Committee of the Consorci Sanitari de Terrassa. It complied with Spanish Law 14/2007 on Biomedical Research and EU Regulation 2016/679 on data protection. The study approval code (CEIm) was 02–23-101-089.

Results
Out of the final sample, 63 incarcerated individuals with cancer were included and compared to 129 community-based patients. Table 1 summarizes the demographic and clinical characteristics of both groups.
The mean age was 58 years (SD ± 13) in the PHU group and 59 years (SD ± 13) in the control group (p > 0.05). Smoking, alcohol consumption, and other substance use were significantly more prevalent in the penitentiary population than in the control population (p < 0.001). In the PHU cohort, 63.5% of patients presented with tumor-related symptoms at the time of diagnosis, while 25.4% exhibited constitutional symptoms such as weight loss, fatigue, or other constitutional symptoms. Interestingly, 14.3% of the cases were diagnosed incidentally during evaluations for unrelated conditions. In the control cohort, 73.0% were diagnosed during hospital admission for tumor-related symptoms, 14.9% with constitutional symptoms, and 3.5% incidentally Table 1.

No significant differences about size were observed for most tumor type, except for prostate cancer, where tumor size was significantly greater in the PHU group (p = 0.033). Statistically significant differences in lymph node involvement were observed in gastrointestinal tumors (62.5% in PHU vs. 29.4% in controls; p = 0.034) and head and neck tumors (84.6% in PHU vs. 69.2% in controls; p = 0.016). Similarly, a higher proportion of head and neck tumors presented with metastatic disease in the PHU group (38.5%) at the time of diagnosis compared to controls (19.2%) (p = 0.034) (Table 2).

At diagnosis, early-stage disease was more frequent in community-based patients (54.7%) than in PHU patients (41.7%), whereas metastatic disease was more common in the PHU cohort (33.3% vs. 23.4%). When grouping early and locally advanced stages versus metastatic stage, the proportions were 66.7% vs. 76.6%, respectively and the difference was not statistically significant (p > 0.05). Regarding the treatments received, no statistically significant differences were found between groups in the use of surgery, chemotherapy, radiotherapy, or biological therapies (Table 3).

Kaplan–Meier survival curves showed no statistically significant differences between the PHU and control groups across tumor types, as assessed by the log-rank test, although the direction and magnitude of the data reveal clinically relevant patterns. Head and neck cancers showed a notably high 10-year mortality in the PHU group (53.8%) compared to 28% in the control group.
Gastrointestinal cancers exhibited similar mortality between groups (26.7% in PHU vs. 32.4% in controls), while prostate and testicular cancers had no recorded deaths in the PHU group over the 10-year follow-up. The most striking difference was observed in lung cancer: 10-year mortality reached 100% in the control group, whereas it was significantly lower in the PHU group (55.6%, p = 0.005) (Table 2; Fig. 2).
Numbers at risk at 2, 5, and 10 years are reported in Fig. 2. The counts decrease markedly over time, with fewer than 10 patients remaining under observation in several subgroups by 5 years and only isolated cases at 10 years.

Discussion
This study reveals that cancer patients in the PHU present a significantly higher prevalence of classic risk factors—such as smoking, alcohol use, and drug consumption—compared to the general population. While this difference has been previously documented [18–21], it not only reflects pre-incarceration health behaviors but also highlights the limitations and challenges faced by preventive health policies during incarceration, especially in settings where structured interventions remain limited [22]. In fact, local data from Catalonia confirm that drug use, including high-risk injection practices, remains prevalent in prison environments, and is associated with marginalization processes within the institution itself [23].
Strikingly, the most frequent tumors among incarcerated patients were gastrointestinal and prostate cancers. This distribution deviates from the general population, where lung, breast, and colorectal cancers are more prevalent [24]. This divergence may reflect limited access to primary and preventive care prior to incarceration, combined with barriers to screening and early detection during privation of liberty. In addition, the burden of chronic infections, malnutrition, and immunological comorbidities in this population may contribute to the development of gastrointestinal and hematologic malignancies [13, 25]. The elevated prevalence of screenable cancers (that is, cancers with organized population screening such as colorectal, breast, or cervical) in this population—particularly among those released from prison within the past 12 months, a period associated with higher care discontinuity—is further exacerbated by systemic barriers to continuity of care, such as fragmented access to medical records, loss of health coverage, and stigmatization of former individuals who are incarcerated, all of which compromise early detection and adherence to treatment [26].
No significant differences were found in the proportion of patients receiving surgery, chemotherapy, or radiotherapy. At first glance, this suggests that individuals who are incarcerated may have had comparable access to oncologic treatments. However, analyzing treatment solely by broad categories fails to capture important nuances, such as treatment intensity, interruptions, toxicity management, or access to supportive therapies. Therefore, while therapeutic access may appear equitable on the surface, doubts remain regarding the actual quality and continuity of care provided [27]. Evidence from studies conducted in England and the United States —health systems with structures and prison healthcare models different from Spain—have documented significant disparities in access to curative treatments among incarcerated individuals, largely due to logistical, security, and administrative barriers [13, 25, 28]. These systemic obstacles have been linked to lower rates of surgical intervention and decreased 5-year survival, even after adjusting for tumor stage and comorbidities [28].
Survival was lower among PHU patients with head and neck cancer, possibly due to later diagnoses, poorer treatment adherence, or logistical barriers to specialized care [29]. Paradoxically, better survival was observed in PHU patients with lung cancer. However, this finding should be interpreted with caution given the small sample size (n = 9), which increases the likelihood of random variation and limits the robustness of the comparison. This unexpected result may also reflect differences in baseline characteristics, intensity of follow-up, or survivorship bias, and therefore warrants cautious interpretation. Moreover, the very limited number of patients remaining under observation beyond five years substantially reduces the reliability of long-term survival estimates and may account for some of the apparent differences observed between groups.
Screening programs in lung cancer have shown survival benefits in high-risk populations through low-dose computed tomography (LDCT) protocols [30, 31]. In Catalonia, however, no organized lung cancer screening program is implemented —only breast, colorectal, and cervical cancer screening programs are available— and systematic LDCT is not performed, as there is no approved protocol from the Catalan Department of Health. Therefore, the better survival observed in PHU lung cancer patients in our cohort cannot be attributed to earlier diagnosis via enhanced surveillance, and our data support this, as the extent of disease at presentation did not differ significantly across groups.
Taken together with the treatment and staging data, these findings suggest that individuals who are incarcerated received cancer care comparable to that of their community-based peers. For prostate and gastrointestinal cancers, survival was similar in both groups, suggesting that once patients are connected to the hospital system, the quality of care provided through the PHU model may match that of the general population [32]. This is particularly relevant, as it suggests that structural integration between hospital and prison units—as exists in our model—may help mitigate health disparities, a scenario rarely seen in other countries [28]. Nevertheless, even in integrated models such as ours, important structural limitations remain that may compromise timely diagnosis, care coordination, and patient outcomes—particularly for those with advanced or complex oncological needs. Other critical domains, such as psychosocial support, palliative care, or post-release follow-up, were not assessed and may still represent significant gaps. Additionally, the small sample sizes within each tumor subgroup limit the statistical power of survival comparisons. As such, these findings should be interpreted with caution and considered hypothesis-generating.
In the prison setting, there are often significant limitations in areas such as timely access to early diagnosis, substantial delays in conducting diagnostic procedures or complementary tests, restrictions or delays in transfers to specialized centers, and logistical barriers that negatively impact both treatment continuity and supportive care (including palliative care, pain management, and psychosocial support). Incarceration itself may therefore not be ethically or logistically compatible with the complex and evolving health needs of patients with chronic and life-threatening conditions such as cancer. In several cohorts, incarceration has been associated with late-stage diagnosis, restricted access to curative care, and significantly higher cancer-related mortality compared to the general population [13, 28, 33]. These findings raise critical questions about the appropriateness of custodial sentences in individuals whose medical condition may render prison settings inadequate or even detrimental to their prognosis.
This discussion is closely linked to broader public health concerns regarding structural determinants of health. The prison population often reflects systemic inequalities, including poverty, housing instability, limited access to healthcare, and a high burden of mental illness—factors that are also associated with increased cancer risk. These social determinants contribute to both incarceration and health disparities. Emerging frameworks such as abolition medicine have called for a shift away from carceral-based health models, advocating instead for equitable care, decarceration policies, and greater investment in community-level health services [26, 34]. These approaches emphasize not only better healthcare within prisons, but also preventive strategies that reduce the risk of these social determinants from vulnerable groups.

Study limitations
This study has several limitations. First, the sample size, particularly the PHU subgroup for some tumor types, is small, which restricts the external validity of our findings. This limitation is conditioned by the inherently small number of individuals who are incarcerated with cancer in the penitentiary system; in fact, all tumors diagnosed in the system have been included since the implementation of electronic medical records. Nonetheless, in fields with scant prior evidence, we believe it is preferable to report all available data rather than suppress potentially informative results.
Second, several control patients were excluded due to incomplete records, which may have introduced modest imbalance between groups. Controls were re-matched 1:2 after exclusions, resulting in 129 controls for 63 PHU cases. Another limitation could be that matching by sex was not performed, as 98.6% of individuals who are incarcerated were male, reflecting the demographic distribution of the prison population. Although this is a descriptive study without a prospective sample size calculation, the chosen 1:2 matching ratio was intended to increase statistical power and improve the reliability of comparisons between the PHU and community groups. This design minimizes selection bias and maximizes the ability to detect statistically significant differences between groups. Estimation of the median survival time requires that at least 50% of patients within a given group experience the event of interest (death). In our cohort, the limited sample size within each tumor-specific subgroup resulted in a low number of deaths, precluding the calculation of a valid median survival. For this reason, we elected not to report median survival times for these groups and acknowledge this as a limitation of our analysis. Moreover, long-term survival estimates must also be interpreted with caution: the number of patients remaining under observation declined substantially over time, with very few individuals evaluable beyond 5 years and only isolated cases at 10 years. These small numbers substantially limit the robustness of survival probabilities, meaning that apparent differences may reflect random variation rather than true population-level effects.
Third, treatment information was recorded only at the level of broad categories (surgery, chemotherapy, radiotherapy, etc.); the absence of regimen-specific details precluded a more granular analysis of therapeutic effects.
Finally, we did not capture comorbidities other than those already reported, so residual confounding by unmeasured health conditions cannot be ruled out. Moreover, the release of patients from the PHU may result in reduced availability of follow-up information once they return to the community, as they may no longer fall under the catchment area of our healthcare hospital.

Conclusions
In our cohort, incarcerated cancer patients showed a higher prevalence of risk factors but received oncological treatments comparable to the general population once admitted to hospital care. Survival analysis revealed poorer outcomes in head and neck cancer and, unexpectedly, better outcomes in lung cancer, although these findings should be interpreted with caution due to the small sample size. The integrated prison–hospital model may help reduce treatment disparities, although barriers to early diagnosis and adequate follow-up persist.

Supplementary Information