Multimorbidity patterns among patients hospitalized with prostate cancer in Portugal: a cluster analysis approach

Multimorbidity patterns among patients hospitalized with prostate cancer in Portugal: a cluster analysis approach | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Multimorbidity patterns among patients hospitalized with prostate cancer in Portugal: a cluster analysis approach Patrícia Carvalho, Julio Souza, Francisco Botelho, Mariana Lobo, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4247648/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Multimorbidity is a common condition among cancer patients, resulting in increased complexity of care and risk of negative outcomes. This study aims to use clustering analysis to identify and characterize multimorbidity patterns among hospitalized prostate cancer patients in Portugal. This is a retrospective observational study using inpatient data from the Portuguese National Hospital Morbidity Database. Data on hospital admissions with a diagnosis of prostate cancer occurring in all public hospitals in mainland Portugal during 2011–2017 were considered. Partitioning clustering algorithms, namely K-modes, PAM (Partitioning Around Medoids), and hierarchical clustering, were used to identify multimorbidity clusters. Results obtained from the different clustering approaches were compared and assessed in terms of clinical relevance. A total of 10394 inpatient episodes were analyzed, with 6091 (58%) reporting multimorbidity. Similar clusters were obtained through the different partitioning approaches, with PAM presenting a higher stability and the best quality results in terms of average silhouette. The analysis of the 6 clusters obtained with PAM reveals groups with a pattern of hypertension co-occurring with diabetes, obesity, and arrhythmia, in addition to cancer itself. In this study, the validity of cluster analysis as an exploratory method for identifying clusters of multimorbid conditions among prostate cancer patients in Portugal was demonstrated, identifying relevant patterns of disease co-occurrence, with potential impact on treatment decisions and outcomes. The identified clusters revealed conditions that typically co-occur with prostate of cancer and that can be controlled throughout all phases of cancer survivorship by means of healthier behaviors aligned with integrated and coordinated care. Clustering Inpatient Prostate Cancer Machine Learning Multimorbidity 1. Introduction Prostate cancer is the most common type of cancer among men and the second most common cause of death from cancer globally [ 1 ]. In 2019, prostate cancer caused 2687 deaths in Portugal and was a cause of premature mortality in that country, accounting for 35.4 Years of Life Lost per 100 000 population, according to the Global Burden of Diseases (GBD) estimates [ 2 ]. In addition to the diagnosis of prostate cancer, a substantial share of the patients presents at least one additional comorbidity, such as diabetes, hypertension and cardiovascular diseases, which in turn is associated with an increased risk of dying from other causes than prostate cancer [ 3 ]. This condition is denominated Multimorbidity, which can be defined as the co-occurrence of two or more chronic conditions [ 4 ] and is becoming increasingly common among hospitalized patients worldwide, adding important challenges for healthcare systems and patients. Regarding cancer care, multimorbidity is of particular concern, as most cancer patients have underlying chronic diseases apart from cancer itself [ 5 ]. Additional comorbidities not only increase the odds of negative outcomes, such as premature death and use of unscheduled hospital services, but also increases the complexity of care [ 6 ]. Literature has identified and documented gaps in existing clinical guidelines for cancer, highlighting the need for greater coordination and integration between cancer and chronic disease care [ 5 ]. Since multimorbidity patients are heterogeneous and can suffer from a wide range of combinations of diseases, overall descriptions of health outcomes and needs based on disease counts are unhelpful for tailoring healthcare actions. Therefore, research aimed at a more tailored understanding of which comorbidities commonly co-occur have been encouraged [ 7 , 8 ]. Our literature search indicates that the increasing interest in multimorbidity has been accompanied by a growing number of published research that investigates this topic through different dimensions: magnitude (e.g., diagnoses count), severity (e.g. Charlson or Elixhauser indexes) or pattern (e.g., clusters of diagnoses) [ 9 ]. Regarding the latter dimension, several studies have recurred to unsupervised machine learning methods, such as cluster analysis, to detect and characterize multimorbidity patterns among cancer patients using hospital administrative data. Some of the most employed methods include Latent Class Analysis [ 10 – 12 ], k-means [ 5 ], k-modes [ 13 ], hierarchical clustering [ 14 – 16 ] and PAM (Partitioning Around Medoids) paired with Gower distance [ 17 ]. No study investigating multimorbidity patterns among prostate cancer hospitalizations has been conducted in the Portuguese context. Therefore, as a first step to understand the implications of multimorbidity clusters on health outcomes such as in-hospital mortality, survival and resource use, we aimed to apply distinct clustering approaches to identify and characterize multimorbidity patterns in a set of inpatient episodes due to prostate cancer using a Portuguese nationwide hospitalization database. 2. Materials and Methods 2.1. Data sources and study population This is retrospective observational study using inpatient data from the Portuguese National Hospital Morbidity Database (NHMD). The NHMD contains data on inpatient and outpatient episodes in all mainland public hospitals within the Portuguese National Health Service. The Central Administration for the Health System (ACSS), which is the health authority responsible for the NHMD, provided access to the data. We restricted the analysis to inpatient episodes with a primary diagnosis of prostate cancer (ICD-9-CM code 185 or ICD-10-CM code C61) and with a discharge date between January 1st 2011 and December 31st 2017. Only inpatient episodes labeled in the database as statistically valid were extracted, that is, those with a length of stay of at least 24h, shorter than 24h for patients who died, left against medical advice, or transferred to another hospital. In Portugal, the two most common usual treatment paths for hospitalized prostate cancer patients are surgery (radical prostatectomy) and brachytherapy. Radical prostatectomy involves the removal of the entire prostate and the seminal vesicles with or without pelvic lymphadenectomy. It can be used for all risk categories of localized prostate cancer. Alternatively, brachytherapy comprises the insertion of radioactive seeds into the prostate gland, being used for men in low and intermediate risk categories of localized prostate cancer [ 18 ]. There are other rarer reasons for hospitalization of a prostatic cancer patient that will not be the focus of this research. Therefore, we further restricted the analysis to inpatient episodes reporting radical prostatectomy (ICD-9-CM procedure code 60.5 or ICD-10-PCS codes 0VT00ZZ and 0VT04ZZ) or brachytherapy (ICD-9-CM procedure codes 92.20, 92.27, 92.28 and 92.29 or ICD-10-PCS codes 0VH031Z and chapter DV10), the two main treatment paths for patients admitted with prostate cancer in Portuguese hospitals. 2.2. Variables As the main goal was to detect patterns of multimorbidity, secondary diagnosis data at episode level was considered to define clusters. These data were transformed into dichotomous variables indicating the presence or absence of Elixhauser comorbidities [ 19 ]. Additionally, dichotomous variables indicating the presence of brachytherapy, surgery and in-hospital death were also considered to form clusters in order to check whether certain multimorbidity patterns may occur along with death or treatment path decision. Moreover, the identified clusters were further interpreted against other variables such as age, variables reflecting resource use, namely length of stay and relative weight, which measures the episode’s cost relative to the average hospitalized national patient, and hospital-related characteristics such as geographic region and hospital group. The latter variable was a category created by ACSS to group hospitals that are similar in terms of case-mix and complexity [ 20 ]. 2.3. Multimorbidity measure Multimorbidity was defined as episodes with ≥ 2 chronic conditions, including the diagnosis of prostate cancer itself. A total of 30 chronic conditions were identified using the Elixhauser method [ 19 ]. These conditions were identified based on a predefined list of ICD-9 and ICD-10 codes that were searched across the secondary diagnosis fields in the HDM data. The full list of specific ICD-9 and ICD-10 codes for the 30 Elixhauser conditions can be found in Quan et al. (2005)[ 21 ]. Cluster analysis was only performed in episodes presenting multimorbidity. 2.4. Cluster analysis Partitioning clustering methods, which are employed to cluster instances based on their level of similarity, can provide a feasible tool for differentiating episodes based on their clinical profile and comorbidity occurrence. The most popular partitioning clustering algorithms is the k-means, but other extensions have been implemented, namely k-modes, and Partitioning Around Medoids (PAM) or CLARA (Clustering Large Applications) algorithms, the latter being an extension of PAM to handle large sample sizes using a sampling approach. [ 22 ]. Since our dataset is only composed of categorical variables, we first tested the k-modes algorithm, which has been widely used to efficiently cluster categorical data [ 23 ]. K-modes addresses the existing limitation of numeric-only data imposed by the k-means algorithm, enabling clustering of large-size categorical data from real-world databases [ 24 ]. It adopts a simple dissimilarity measurement adjusted for categorical data denominated total within-cluster simple-matching distance, which compares two instances and computes the number of matching variables relative to the total number of variables. In addition to K-modes, PAM and CLARA also extend the k-means clustering algorithm, being less sensitive to outliers and suitable for categorical variables. These approaches find instances in each cluster as a representative object (medoids) and assign all the instances to the clusters they best fit in. PAM starts by finding a medoid, which should be the most centrally located observation in the cluster. Following the determination of the initial set of medoids, one medoid is iteratively replaced by one non-medoid whenever the total distance of the resulting clustering is improved [ 22 ]. PAM was the preferable method over CLARA considering the study’s sample size and computing time. Apart from testing the partitioning methods, hierarchical clustering was also tested. This method is widely used in several fields and produces a hierarchy of clusters represented in a tree-like structure called dendrogram, where it is possible to identify where clusters split and what sub-clusters are subsequently created [ 25 ]. Hierarchical clustering methods are often regarded as good quality clustering methods, creating clusters of arbitrary shape and are generally robust to noise. There are several approaches to apply hierarchical clustering in the literature. In this study, the agglomerative approach was considered. This approach implements a bottom-up method that starts with all instances forming a separate cluster before being iteratively merged into similarity clusters. In this process, Ward’s minimum variance was considered. Ward’s method aims at minimizing the total within-cluster variance, in which the pair of clusters with minimum between-cluster distance are merged. Moreover, Elbow’s method and Silhouette analysis [ 26 ] were used to find the optimal number of clusters k. Regarding dissimilarity measurement, k-modes was run considering the total within-cluster simple-matching distance, whereas PAM and Hierarchical clustering were run on a Gower distance matrix calculated from the data, which, unlike other commonly used distance metrics, can handle mixed types of data. 2.5. Statistical analysis Baseline characteristics of the included population were summarized by mean, median and interquartile range (IQR) for continuous variables, and frequencies (n) and percentages for categorical variables. Moreover, the Observed to Expected (O/E) prevalence ratios of Elixhauser comorbidities were computed by dividing the prevalence of each comorbidity within each identified cluster by the corresponding prevalence of that condition in the entire analyzed population. Following the approach of Jansana et al. (2021)[ 16 ], a chronic condition was highlighted within a cluster whenever its O/E prevalence ratio was at least 2. All statistical analyses were performed using R language and RStudio software. The R package “icd” was used to identify the Elixhauser conditions in the data according to Quan et al. (2005)[ 21 ] definitions. K-modes clustering required the package “klaR”, whereas PAM and hierarchical clustering required packages “cluster” and “stats”. 3. Results The analyzed dataset included a total of 10394 inpatient episodes, with most of them performing radical prostatectomy (n = 8565, 82.0%) (Table 1 ). Nearly half of the included episodes reported multimorbidity (n = 6286, 58%), with the most prevalent comorbidities being hypertension (n = 4596, 44.0%), diabetes mellitus (n = 1413, 14.0%) and obesity (n = 826, 7.9%). Additionally, a total of 336 episodes (3.2% of the total) unexpectedly reported metastasis and a very low frequency of in-hospital death (n = 31, 0.3%) was observed. The median age was 65 years old (IQR: 60–69) and the median length of stay was 5 days (IQR:4–7). Most episodes occurred in the Northern region (n = 4215, 41%), followed by Lisbon (n = 3050, 30%) and Centre (n = 2659, 26%). Hospitals belonging to group E, which is composed of the largest facilities with teaching status, admitted the highest number of episodes (n = 3318, n = 32%), followed by group F (n = 2678, 26%), composed of oncology hospitals, and group C (n = 2172, 21%), composed of medium-sized hospitals. Table 1 Descriptive Statistics of the analyzed dataset Characteristics Overall Age 1 65 (60–69) Brachytherapy 2 1830 (18%) Radical Prostatectomy 2 8565 (82%) Metastasis 2 336 (3.2%) Length of stay 1 5 (4–7) Death 2 31 (0.3%) ‍Multimorbidity 2 6286 (58%) 1. Median (IQR); 2. Number of episodes (share of the total) Table 2 presents the clusters obtained with the different clustering approaches, including selected cluster quality metrics. Each cluster was presented according to the mode of the most prevalent comorbidity or treatment performed. Thus, comorbidity names appearing in the designation of the clusters occurred in at least 50% of the episodes within the clusters. The optimal number of clusters chosen for k-modes was 4, whereas the optimal number of clusters chosen for PAM and hierarchical clustering were 6 and 8, respectively. Clusters obtained with all three methods identified groups with high co-occurrence of hypertension paired with diabetes. However, the K-modes algorithm performed poorly compared to the other approaches as it provided the lowest silhouette scores (Table 2 ). Additionally, K-modes was more unstable, meaning that repeated experiments generated different clusters, apart from producing redundant clusters. Table 2 Identified multimorbidity clusters among prostate cancer hospitalizations using different clustering approaches. Method Clusters Silhouette index K-modes Cluster 1 (n = 3690) - Hypertension (69%) and Surgery (89%) Cluster 2 (n = 628) - Hypertension (100%), Diabetes (86%), Surgery (86%) Cluster 3 (n = 882) - Hypertension (62%), brachytherapy (58%) Cluster 4 (n = 891) - Hypertension (100%), Surgery (88%) -0.09 0.67 -0.24 0.08 Average score: − 0.006 PAM Cluster 1 (n = 2839) - Hypertension (84%), Surgery (100%) Cluster 2 (n = 921) - Hypertension (75%), Diabetes Mellitus (100%), Surgery (100%) Cluster 3 (n = 366) - Hypertension (66%), Arrhythmia (100%), Surgery (100%) Cluster 4 (n = 620) - Hypertension (72%), Obesity (100%), Surgery (100%) Cluster 5 (n = 265) - Chronic pulmonary disease (100%), Surgery (82%) Cluster 6 (n = 1080) - Hypertension (78%), Brachytherapy (100%) 0.49 0.58 0.25 0.30 0.37 0.38 Average score: 0.45 Hierarchical Cluster 1 (n = 2577) - Hypertension (86%), Surgery (100%) Cluster 2 (n = 1008) - Hypertension (76%), Diabetes Mellitus (100%), Surgery (100%) Cluster 3 (n = 170) - Depression (100%), Surgery (100%) Cluster 4 (n = 319) - Hypertension (64%), Arrhythmia (100%), Surgery (100%) Cluster 5 (n = 336) - Chronic pulmonary disease (100%), Surgery (100%) Cluster 6 (n = 364) - Hypertension (65%), Obesity (100%), Surgery (100%) Cluster 7 (n = 189) - Alcohol abuse (89%), Surgery (100%) Cluster 8 (n = 1128) - Hypertension (75%), Brachytherapy (100%) 0.46 0.44 0.54 0.27 0.44 0.64 0.15 0.33 Average score: 0.41 The use of hierarchical and PAM approaches tended to produce more stable and better-quality clusters, with groups presenting clearly defined patterns in terms of comorbidity co-occurrence. Apart from the usual co-occurrence of prostate cancer with hypertension, either among surgical or brachytherapy cases, both methods found three clusters with high co-occurrence of hypertension with diabetes, obesity, and arrhythmia, apart from clusters with high co-occurrence of cancer and chronic pulmonary disease among surgical cases (Table 2 ). Moreover, hierarchical clustering detected clusters of surgical episodes with high co-occurrence of prostate cancer with depression (Cluster 3) or alcohol abuse (Cluster 7). Since the average silhouette was higher for the clusters obtained with the PAM method, we presented the assessment of the clusters obtained with the PAM method against age, resource use and hospital-related variables (Table 3 ). Only Elixhauser comorbidities with an overall prevalence above 1% are presented in Table 3 and highlighted prevalences of comorbidities are related to a O/E ratio higher than 2. Table 3 Multimorbidity clusters obtained with the PAM approach and descriptions based on the prevalence of selected Elixhauser comorbidities and other episode characteristics. Variables Cluster 1 (n = 2839) Cluster 2 (n = 921) Cluster 3 (n = 366) Cluster 4 (n = 620) Cluster 5 (n = 265) Cluster 6 (n = 1080) Total (n = 10394) No multimorbidity (n = 4303) Arrhythmia, n(%) 0 (0) 0 (0) 366 (100) 0 (0.0) 17 (6.4) 125 (12) 508 (4.9) - Hypertension, n(%) 2377 (84) 688 (75) 242 (66) 445 (72) 0 (0.0) 844 (78) 4596 (44) - Chronic pulmonary disease, n(%) 151 (5.3) 40 (4.3) 21 (5.7) 40 (6.5) 265 (100) 64 (5.9) 581 (5.6) - Diabetes Mellitus, n(%) 0 (0.0) 921 (100) 53 (14) 161 (26) 20 (7.5) 258 (24) 1413 (14) - Renal disease, n(%) 85 (3.0) 27 (2.9) 13 (3.6) 13 (2.1) 5 (1.9) 29 (2.7) 172 (1.7) - Peptic ulcer disease, n(%) 61 (2.1) 11 (1.2) 3 (0.8) 4 (0.6) 5 (1.9) 18 (1.7) 102 (1.0) - Other Tumors, n(%) 105 (3.7) 15 (1.6) 6 (1.6) 16 (2.6) 7 (2.6) 39 (3.6) 188 (1.8) - Obesity, n(%) 0 (0) 0 (0) 45 (12) 620 (100) 28 (11) 133 (12) 826 (7.9) - Alcohol abuse, n(%) 141 (5.0) 29 (3.1) 13 (3.6) 38 (6.1) 23 (8.7) 140 (13) 384 (3.7) - Depression, n(%) 169 (6.0) 20 (2.2) 11 (3.0) 22 (3.5) 11 (4.2) 55 (5.1) 288 (2.8) - Metastasis, n(%) 78 (2.7) 29 (3.1) 9 (2.5) 17 (2.7) 8 (3.0) 59 (5.5) 336 (3.2) 136 (3.2) Comorbidity count, median (IQR) 1 (1–1) 2 (2–2) 2 (2–3) 2 (2–3) 1 (1–2) 1 (1–2) 1 (0–2) - Age, median (IQR) 65 (61–69) 65 (62–69) 66 (62–70) 64 (60–68) 65 (60–68) 67 (62–71) 65 (60–69) 64 (59–68) Length of stay, median (IQR) 6 (4–7) 6 (4–7) 6 (4–7) 5 (4–7) 5 (3–7) 2 (2–3) 5 (4–7) 5 (4, 7) Relative weight, mean (standard deviation) 1.13 (± 0.74) 1.19 (± 0.89) 1.30 (± 0.86) 1.17 (± 0.85) 2.57 (± 2.95) 8.50 (± 1.25) 6.36 (± 1.71) 5.35 (± 1.56) Brachytherapy, n(%) 0 (0) 0 (0) 0 (0) 0 (0.0) 49 (18) 1080 (100) 1830 (18) 701 (16) Surgery, n(%) 2,839 (100) 921 (100) 366 (100) 620 (100) 216 (82) 0 (0.0) 8565 (82) 3603 (84) In-hospital death, n(%) 1 (< 0.1) 0 (0.0) 4 (1.1) 1 (0.2) 0 (0.0) 19 (1.8) 31 (0.3) 6 (0.1) Hospital group, mode (%) Group E (37) Group E (40) Group E (41) Group E (37) Group E (29) Group F (88) Group E (32) Group E (32) Geographic region, mode (%) North (44) North (40) North (41) North (37) North (46) Centre (53) North (41) North (41) Cluster 6 concentrated nearly all brachytherapy episodes, characterized mostly by the high co-occurrence of hypertension and cancer, although it also presented a O/E ratio of alcohol abuse higher than 2. Also, this cluster presented a substantially higher cost in terms of relative weights when compared to other clusters, despite being composed only of brachytherapy episodes. Accordingly, Cluster 6 presented the highest proportions of in-hospital death and metastasis, suggesting increased severity. Similar to Cluster 6, Cluster 5 also presents a O/E ratio of alcohol abuse higher than 2, with the difference that it comprises mainly surgical episodes presenting high co-occurrence of cancer with chronic pulmonary disease rather than hypertension. Cluster 1 presented an O/E ratio above 2 for peptic ulcer disease, other tumors, and depression, possibly co-occurring with hypertension. Cluster 3, which is only composed of prostate cancer episodes co-occurring with arrhythmia, presented an O/E ratio above 2 for renal disease. Regarding hospital-related characteristics, the North region was the most representative region across all clusters, except for Cluster 6, which was mostly composed of episodes admitted in hospitals from the Centre region. Furthermore, unlike the observed for other clusters, which were mostly represented by group E hospitals, Cluster 6 concentrated mainly episodes admitted into group F hospitals. Despite most clusters presenting a very high prevalence of hypertension, none reported an O/E ratio below 2 for this condition due to the high prevalence in the overall included population. Finally, as expected, clusters presenting surgery (radical prostatectomy) reported a higher length of stay. No major differences concerning median age and comorbidity count were observed across the six clusters. 4. Discussion To the best of our knowledge, this is the first study to use clustering analysis to assess multimorbidity patterns in prostate cancer hospitalizations in Portugal, presenting an almost nationwide nature, with a large number of episodes. Different clustering approaches were used, with PAM and hierarchical clustering producing the best results, with similar clusters in terms of main clinical characteristics. These two clustering algorithms were useful and practical approaches to detect multimorbidity patterns, as they clearly differentiated episodes according to the two main treatment paths for hospitalized patients in Portugal (brachytherapy and radical prostatectomy), apart from revealing prevalent co-occurrences of diseases with potential impact on cancer survival and on the treatment itself. Nearly half (58%) of the analyzed prostate cancer episodes reported multimorbidity. According to Rizzuto et al. (2017) [ 27 ], the relationship between multimorbidity and mortality has been extensively studied, however the conclusions have been inconsistent, with some studies reporting a negative association between the presence of multiple comorbidities and survival, whereas others found no association. Considering the negligible number of in-hospital deaths and the pseudonymized nature of our data, which does not allow longitudinal tracking of individuals, we could not find any relationship between multimorbidity and mortality or survival. Furthermore, contrary to previous evidence documented elsewhere in the literature [ 28 ], our results did not find a relationship between the multimorbidity and length of stay. The lower length of stay observed for Cluster 6 (Table 3 ) was likely to be attributed to the concentration of brachytherapy episodes, whether the remaining clusters were essentially composed of surgical episodes. Six clusters identified with the PAM method were chosen for a more detailed interpretation (Table 3 ). The detected clusters presented a similar number of comorbidities, with the highest number consisting of two chronic diseases in addition to prostate cancer (Clusters 2, 3 and 4). Regarding the main patterns observed, Cluster 1 and Cluster 6 were composed exclusively of episodes with a high co-occurence of cancer with hypertension for episodes reporting surgery and brachytherapy, respectively, although Cluster 6 presented the highest burden in terms of costs. Although a clear explanation could not be immediately found, this cluster accounted for the highest prevalence of mortality and metastasis, suggesting that it concentrated more episodes with increased severity and risk of mortality in comparison with the remaining clusters, resulting in higher inpatient costs. Clusters 2, 3 and 4 were composed of surgical episodes reporting a high co-occurrence of hypertension with diabetes mellitus, arrhythmia, and obesity, respectively, in addition to prostate cancer, which was also found using the hierarchical clustering method. In Cluster 5, chronic pulmonary disease co-occurred mostly with prostate cancer alone. The coexistence of several conditions from different medical specialties in the same cluster constitutes a challenging situation for doctors and patients. Although previous literature suggests that integrated care for multimorbid patients can improve health outcomes and efficiency of care, evidence is still limited and may only be feasible in countries with relatively strong and well-resourced health systems [ 29 ]. These results are also aligned with previous literature, in which prostate cancer diagnosis co-occurred mostly with comorbidities such as hypertension and diabetes, either for localized or advanced patients [ 5 , 30 ]. Moreover, the identified clusters were composed of episodes reporting conditions that may be associated with modifiable risk factors with potential impact on the risk of developing or dying from prostate cancer, such as high fat diet and smoking [ 31 , 32 ]. However, as there is no sufficient evidence on clear indications for prevention beyond early diagnosis to reduce prostate cancer mortality [ 33 ], the focus should be on increasing efforts to manage chronic diseases, namely hypertension and diabetes, during cancer treatment, which includes efforts to control behaviors throughout all phases of prostate cancer survivorship. The identified clusters are clinically relevant in a sense they reveal potential risk groups that should be targeted by different levels of care in order to ensure greater survival and avoid non-planned visits to the health services. The geographic distribution of patients found across the clusters corroborates with the National Oncological Registry, which identified regional variations in the incidence of prostate cancer in Portugal in 2018, with the highest incidences in Lisbon, North and Centre regions [ 34 ]. These regions should thereby be the focus of efforts aimed at implementing oncology referral networks, reinforcing the importance of multidisciplinary teams that can provide integrated care to multimorbid, difficult-to-treat patients. Regarding the treatment path, Cluster 6 suggests an asymmetric distribution of brachytherapy across hospital groups and geographical regions, as most patients were admitted in the Centre region and mostly in oncology hospitals, possibly indicating different treatment practices between regions and hospital types. It is important to state the underlying limitations of this study. First, hospital data used for the analyses was obtained from an administrative database, which in turn presents underlying quality issues, mostly related to clinical coding and the degree to which comorbidities are documented and coded [ 35 ]. Nevertheless, in Portugal, clinical coding is exclusively performed by medical doctors with specific training and verified by internal and external audits. Moreover, as data was de-identified, individual patients could not be identified nor followed. Furthermore, the clustering approaches considered in this study do not consider the temporal changes in the patients’ condition, and episodes are restricted and assumed to be part of a single cluster, rather than the possibility of belonging to more than one cluster simultaneously. Conclusions Cluster analysis is a useful approach to detect and characterize different patterns and profiles of prostate cancer hospitalizations in Portugal. Multimorbidity is present in 58% of the admissions for prostate cancer treatment, with these additional conditions potentially accounting for their cause of death, thereby highlighting the importance of addressing the comorbidity status by integrating and coordinating chronic disease care along with cancer care, whenever possible. In particular, the clusters revealed the existence of well-defined and distinct multimorbidity profiles, in which cancer co-occurs mostly with hypertension and other additional chronic disease, such as diabetes, obesity and arrhythmia, potentially incurring a higher treatment burden, requiring thus increased attention during treatment. Future studies should ideally include the assessment of the association between the different identified clusters and survival, recurring to record linkage techniques. Declarations Ethics approval and consent to participate Data access was granted by the Central Authority for Health Services, I.P. (ACSS) through a research protocol with the Faculty of Medicine of the University of Porto. Since inpatient data used in this study were previously de-identified by ACSS, there was no need for ethical approval. Consent for publication Not applicable Availability of data and materials The data that support the findings of this study are available from the Central Authority for Health Services, I.P. (ACSS). Restrictions apply to the availability of these data, which were used under license for this study. Data are however available from the authors upon reasonable request and with permission of ACSS. Competing interests The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Funding This research was funded by Norte2020 (Portugal 2020) within the project Sex Health & Prostate Cancer - Biopsychological Determinants of Sexual Health in Men with Prostate Cancer (NORTE-01-0145-FEDER-000057). It has received Portuguese National Funds through FCT (Portuguese Foundation for Science and Technology) under project UIDB/00760/2020 of the GECAD. Authors’ contributions A. F. was responsible for the conceptualization; A.F. contributed to the data access; P.C. performed literature review; all authors contributed to the methodology design; P.C. and J.S. conducted data analysis; P.C., J.S., M.L. and F.B. wrote the main manuscript text; J.S. prepared tables 1-3; all authors reviewed the manuscript. Acknowledgments The authors would like to thank the Central Authority for Health Services, I.P. (ACSS) for providing access to the data. Sex Health & Prostate Cancer - Biopsychological Determinants of Sexual Health in Men with Prostate Cancer (NORTE-01-0145-FEDER-000057). It has received Portuguese National Funds through FCT (Portuguese Foundation for Science and Technology) under project UIDB/00760/2020 of the GECAD. References Silva Gaspar, S. R., Fernandes, M., Castro, A., Oliveira, T., Santos Dias, J., & Palma Dos Reis, J. Active surveillance protocol in prostate cancer in Portugal. Actas Urológicas Españolas (English Edition) 46(6), 329–339 (2022). https://doi.org/10.1016/j.acuroe.2022.01.002 Global burden of disease study 2019 (GBD 2019) data resources. Healthdata.org., https://ghdx.healthdata.org/gbd-2019, last accessed 2023/04/14 Jefferson, M., Drake, R. R., Lilly, M., Savage, S. J., Tucker Price, S., & Hughes Halbert, C. Co-morbidities in a Retrospective Cohort of Prostate Cancer Patients. Ethnicity & disease, 30(Suppl 1), 185–192 (2020). https://doi.org/10.18865/ed.30.S1.185 Pefoyo, A. J. K., Bronskill, S. E., Gruneir, A., Calzavara, A., Thavorn, K., Petrosyan, Y., Maxwell, C. J., Bai, Y., & Wodchis, W. P. The increasing burden and complexity of multimorbidity. BMC Public Health, 15(1), 415 (2015). https://doi.org/10.1186/s12889-015-1733-2 Koné, A. P., & Scharf, D. Prevalence of multimorbidity in adults with cancer, and associated health service utilization in Ontario, Canada: a population-based retrospective cohort study. BMC cancer, 21(1), 406 (2021). https://doi.org/10.1186/s12885-021-08102-1 The Academy of Medical Sciences. Multimorbidity: A Priority for Global Health Research, https://acmedsci.ac.uk/policy/policy-projects/multimorbidity, last accessed 2023/04/03. The Academy of Medical Sciences. Multimorbidity: A Priority for Global Health Research. https://acmedsci.ac.uk/policy/policy-projects/multimorbidity, last accessed 2023/08/25. Medical Research Council. Multimorbidity in the UK Population: Understanding Disease Clustering. 2018; https://mrc.ukri.org/funding/browse/multimorbidity/multimorbidity-in-the-uk-population-understanding-disease-clustering/, last accessed 2023/08/25. Haregu T., Oldenburg B., Setswe G., Elliott J. Perspectives, constructs and methods in the measurement of multimorbidity and comorbidity: A critical review. The Internet Journal of Epidemiology, 10(2) (2012). https://doi.org/10.5580/2cc3 Harrington, R. L., Qato, D. M., Antoon, J. W., Caskey, R. N., Schumock, G. T., & Lee, T. A. Impact of multimorbidity subgroups on the health care use of early pediatric cancer survivors. Cancer, 126(3), 649–658 (2020). https://doi.org/10.1002/cncr.32201 Fillmore, N. R., DuMontier, C., Yildirim, C., La, J., Epstein, M. M., Cheng, D., Cirstea, D., Yellapragada, S., Abel, G. A., Gaziano, J. M., Do, N., Brophy, M., Kim, D. H., Munshi, N. C., & Driver, J. A. Defining multimorbidity and its impact in older United States veterans newly treated for multiple myeloma. Journal of the National Cancer Institute, 113(8), 1084–1093 (2021). https://doi.org/10.1093/jnci/djab007 McLoone, P., Jani, B. D., Siebert, S., Morton, F. R., Canning, J., Macdonald, S., Mair, F. S., & Nicholl, B. I. Classification of long-term condition patterns in rheumatoid arthritis and associations with adverse health events: a UK Biobank cohort study. Journal of Multimorbidity and Comorbidity, 13, 263355652211486 (2023). https://doi.org/10.1177/26335565221148616 Bucholc, M., Bradley, D., Bennett, D., Patterson, L., Spiers, R., Gibson, D., Van Woerden, H., & Bjourson, A. J. Identifying pre-existing conditions and multimorbidity patterns associated with in-hospital mortality in patients with COVID-19. Scientific Reports, 12(1), 17313 (2022). https://doi.org/10.1038/s41598-022-20176-w Lai, F. T. T., Beeler, P. E., Yip, B. H. K., Cheetham, M., Chau, P. Y. K., Chung, R. Y., Wong, E. L. Y., Yeoh, E.-K., Battegay, E., & Wong, S. Y. S. Comparing multimorbidity patterns among discharged middle-aged and older inpatients between Hong Kong and Zurich: A hierarchical agglomerative clustering analysis of routine hospital records. Frontiers in Medicine, 8, 651925 (2021). https://doi.org/10.3389/fmed.2021.651925 Cezard, G., Sullivan, F., & Keenan, K. Understanding multimorbidity trajectories in Scotland using sequence analysis. Scientific Reports, 12(1), 16485 (2022). https://doi.org/10.1038/s41598-022-20546-4 Jansana, A., Poblador-Plou, B., Gimeno-Miguel, A., Lanzuela, M., Prados-Torres, A., Domingo, L., Comas, M., Sanz-Cuesta, T., Del Cura-Gonzalez, I., Ibañez, B., Abizanda, M., Duarte-Salles, T., Padilla-Ruiz, M., Redondo, M., Castells, X., Sala, M., & SURBCAN Group. Multimorbidity clusters among long-term breast cancer survivors in Spain: Results of the SURBCAN study. International Journal of Cancer. Journal International Du Cancer, 149(10), 1755–1767 (2021). https://doi.org/10.1002/ijc.33736 Robertson, L., Vieira, R., Butler, J., Johnston, M., Sawhney, S., & Black, C. Identifying multimorbidity clusters in an unselected population of hospitalised patients. Scientific Reports, 12(1), 5134 (2022). https://doi.org/10.1038/s41598-022-08690-3 Mottet N., van den Bergh R.C.N., Briers E., Van den Broeck T., Cumberbatch M.G., De Santis M., Fanti S., Fossati N., Gandaglia G., Gillessen S., Grivas N., Grummet J., Henry A.M., van der Kwast T.H., Lam T.B., Lardas M., Liew M., Mason M.D., Moris L., Oprea-Lager D.E., van der Poel H.G., Rouvière O., Schoots I.G., Tilki D., Wiegel T., Willemse P.M., Cornford P. EAU-EANM-ESTRO-ESUR-SIOG Guidelines on Prostate Cancer-2020 Update. Part 1: Screening, Diagnosis, and Local Treatment with Curative Intent. European urology, 79(2), 243–262 (2021). https://doi.org/10.1016/j.eururo.2020.09.042 Elixhauser, A., Steiner, C., Harris, D. R., & Coffey, R. M. Comorbidity measures for use with administrative data. Medical Care, 36(1), 8–27 (1998). https://doi.org/10.1097/00005650-199801000-00004 Souza, J., Caballero, I., Vasco Santos, J., Lobo, M., Pinto, A., Viana, J., Sáez, C., Lopes, F., & Freitas, A. Multisource and temporal variability in Portuguese hospital administrative datasets: Data quality implications. Journal of Biomedical Informatics, 136(104242), 104242 (2022). https://doi.org/10.1016/j.jbi.2022.104242 Quan, H., Sundararajan, V., Halfon, P., Fong, A., Burnand, B., Luthi, J. C., Saunders, L. D., Beck, C. A., Feasby, T. E., & Ghali, W. A. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Medical care, 43(11), 1130–1139 (2005). https://doi.org/10.1097/01.mlr.0000182534.19832.83 Hao, M., Li, Y., & Yamamoto, T. Public preferences and willingness to pay for shared autonomous vehicles services in Nagoya, Japan. Smart Cities, 2(2), 230–244 (2019). https://doi.org/10.3390/smartcities2020015 Zhou, H., Zhang, Y., & Liu, Y. A Global-Relationship Dissimilarity Measure for the k-Modes Clustering Algorithm. Computational intelligence and neuroscience, 2017, 3691316 (2017). https://doi.org/10.1155/2017/3691316 Bai, L., & Liang, J. The k-modes type clustering plus between-cluster information for categorical data. Neurocomputing, 133, 111–121 (2014). https://doi.org/10.1016/j.neucom.2013.11.024 Johnson, S. C. Hierarchical clustering schemes. Psychometrika, 32(3), 241–254 (1967). https://doi.org/10.1007/bf02289588 Rousseeuw, P. J. Silhouettes: A graphical aid to the interpretation and validation of cluster analysis. Journal of Computational and Applied Mathematics, 20, 53–65 (1987). https://doi.org/10.1016/0377-0427(87)90125-7 Rizzuto, D., Melis, R. J. F., Angleman, S., Qiu, C., & Marengoni, A. Effect of chronic diseases and multimorbidity on survival and functioning in elderly adults. Journal of the American Geriatrics Society, 65(5), 1056–1060 (2017). https://doi.org/10.1111/jgs.14868 Bayer-Oglesby, L., Zumbrunn, A., Bachmann, N., & on behalf of the SIHOS Team. Social inequalities, length of hospital stay for chronic conditions and the mediating role of comorbidity and discharge destination: A multilevel analysis of hospital administrative data linked to the population census in Switzerland. PloS One, 17(8), e0272265 (2022). https://doi.org/10.1371/journal.pone.0272265 Navickas, R., Petric, V. K., Feigl, A. B., & Seychell, M. Multimorbidity: What do we know? What should we do?. Journal of comorbidity, 6(1), 4–11 (2016). https://doi.org/10.15256/joc.2016.6.72 Benzo, R., Moreno, P. I., Fox, R. S., Silvera, C. A., Walsh, E. A., Yanez, B., Balise, R. R., Oswald, L. B., & Penedo, F. J. Comorbidity burden and health-related quality of life in men with advanced prostate cancer. Research square, rs.3.rs-2572781 (2023). https://doi.org/10.21203/rs.3.rs-2572781/v1 Adjakly, M., Ngollo, M., Dagdemir, A., Judes, G., Pajon, A., Karsli-Ceppioglu, S., Penault-Llorca, F., Boiteux, J. P., Bignon, Y. J., Guy, L., & Bernard-Gallon, D. Prostate cancer: The main risk and protective factors-Epigenetic modifications. Annales d'endocrinologie, 76(1), 25–41 (2015). https://doi.org/10.1016/j.ando.2014.09.001 Brookman-May, S. D., Campi, R., Henríquez, J. D. S., Klatte, T., Langenhuijsen, J. F., Brausi, M., Linares-Espinós, E., Volpe, A., Marszalek, M., Akdogan, B., Roll, C., Stief, C. G., Rodriguez-Faba, O., & Minervini, A. Latest Evidence on the Impact of Smoking, Sports, and Sexual Activity as Modifiable Lifestyle Risk Factors for Prostate Cancer Incidence, Recurrence, and Progression: A Systematic Review of the Literature by the European Association of Urology Section of Oncological Urology (ESOU). European urology focus, 5(5), 756–787 (2019). https://doi.org/10.1016/j.euf.2018.02.007 Gandaglia, G., Leni, R., Bray, F., Fleshner, N., Freedland, S. J., Kibel, A., Stattin, P., Van Poppel, H., & La Vecchia, C. Epidemiology and Prevention of Prostate Cancer. European urology oncology, 4(6), 877–892 (2021). https://doi.org/10.1016/j.euo.2021.09.006 Ministry of Health (2018). Registo oncológico nacional de todos os tumores na população residente de Portugal. Retrieved from: https://ron.min-saude.pt/media/2196/2021-0518_publica%C3%A7%C3%A3o-ron_2018.pdf, last accessed 23/08/18 Souza, J., Santos, J. V., Canedo, V. B., Betanzos, A., Alves, D., & Freitas, A. (2020). Importance of coding co-morbidities for APR-DRG assignment: Focus on cardiovascular and respiratory diseases. Health information management: journal of the Health Information Management Association of Australia, 49(1), 47–57 (2020). https://doi.org/10.1177/1833358319840575 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4247648","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":290470856,"identity":"9b2ced9f-794f-48b0-97e4-5c32c070bb3d","order_by":0,"name":"Patrícia Carvalho","email":"","orcid":"","institution":"Institute of Engineering – Polytechnic of Porto","correspondingAuthor":false,"prefix":"","firstName":"Patrícia","middleName":"","lastName":"Carvalho","suffix":""},{"id":290470857,"identity":"b56c225a-9468-4f41-be52-9224cd37cdb2","order_by":1,"name":"Julio Souza","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAu0lEQVRIiWNgGAWjYBACxhkMjA8SQCxmErQwGyBpMSBCjwQDmwQSlwgtzLPbn1U8+GOTx8DOe+wB444/RDhszhmzG4ltacUMzHzpBoxniLCFcUYO243EhsOJDcw8ZhKMbURpSX9WkPDnP0laEswYEtgOkKJlzhljicS25GI2kF8SzxgT1mI4u/3hxx9/7PL4+c8ee/BxhxwRWhogdAIbAw8bQ2IDYR0M8lA6gQGkhZEYLaNgFIyCUTDiAAC7ITUbPTIOiQAAAABJRU5ErkJggg==","orcid":"","institution":"Institute of Engineering – Polytechnic of Porto","correspondingAuthor":true,"prefix":"","firstName":"Julio","middleName":"","lastName":"Souza","suffix":""},{"id":290470858,"identity":"20d8fa6d-5681-431b-bdda-63f744045cf5","order_by":2,"name":"Francisco Botelho","email":"","orcid":"","institution":"Centro Hospitalar e Universitário de São João","correspondingAuthor":false,"prefix":"","firstName":"Francisco","middleName":"","lastName":"Botelho","suffix":""},{"id":290470859,"identity":"fc5b69bd-bd38-40fc-ab3c-3be3ce5ddd05","order_by":3,"name":"Mariana Lobo","email":"","orcid":"","institution":"University of Porto","correspondingAuthor":false,"prefix":"","firstName":"Mariana","middleName":"","lastName":"Lobo","suffix":""},{"id":290470860,"identity":"f2792e1a-5810-4fb5-8ddb-a8fca68cca75","order_by":4,"name":"Goreti Marreiros","email":"","orcid":"","institution":"Institute of Engineering – Polytechnic of Porto","correspondingAuthor":false,"prefix":"","firstName":"Goreti","middleName":"","lastName":"Marreiros","suffix":""},{"id":290470861,"identity":"fe01c4e8-03cd-497d-8675-6652587ed342","order_by":5,"name":"Alberto Freitas","email":"","orcid":"","institution":"University of Porto","correspondingAuthor":false,"prefix":"","firstName":"Alberto","middleName":"","lastName":"Freitas","suffix":""}],"badges":[],"createdAt":"2024-04-10 13:29:42","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4247648/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4247648/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":66511581,"identity":"65a3c34a-f3ac-4f4b-afdd-de1ac6e12aba","added_by":"auto","created_at":"2024-10-13 23:08:27","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":620351,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4247648/v1/1ebc33ef-698c-4f7b-a5b9-5b6f48e3e68a.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Multimorbidity patterns among patients hospitalized with prostate cancer in Portugal: a cluster analysis approach","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eProstate cancer is the most common type of cancer among men and the second most common cause of death from cancer globally [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. In 2019, prostate cancer caused 2687 deaths in Portugal and was a cause of premature mortality in that country, accounting for 35.4 Years of Life Lost per 100 000 population, according to the Global Burden of Diseases (GBD) estimates [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. In addition to the diagnosis of prostate cancer, a substantial share of the patients presents at least one additional comorbidity, such as diabetes, hypertension and cardiovascular diseases, which in turn is associated with an increased risk of dying from other causes than prostate cancer [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. This condition is denominated Multimorbidity, which can be defined as the co-occurrence of two or more chronic conditions [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e] and is becoming increasingly common among hospitalized patients worldwide, adding important challenges for healthcare systems and patients.\u003c/p\u003e \u003cp\u003eRegarding cancer care, multimorbidity is of particular concern, as most cancer patients have underlying chronic diseases apart from cancer itself [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Additional comorbidities not only increase the odds of negative outcomes, such as premature death and use of unscheduled hospital services, but also increases the complexity of care [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Literature has identified and documented gaps in existing clinical guidelines for cancer, highlighting the need for greater coordination and integration between cancer and chronic disease care [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Since multimorbidity patients are heterogeneous and can suffer from a wide range of combinations of diseases, overall descriptions of health outcomes and needs based on disease counts are unhelpful for tailoring healthcare actions. Therefore, research aimed at a more tailored understanding of which comorbidities commonly co-occur have been encouraged [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eOur literature search indicates that the increasing interest in multimorbidity has been accompanied by a growing number of published research that investigates this topic through different dimensions: magnitude (e.g., diagnoses count), severity (e.g. Charlson or Elixhauser indexes) or pattern (e.g., clusters of diagnoses) [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Regarding the latter dimension, several studies have recurred to unsupervised machine learning methods, such as cluster analysis, to detect and characterize multimorbidity patterns among cancer patients using hospital administrative data. Some of the most employed methods include Latent Class Analysis [\u003cspan additionalcitationids=\"CR11\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e], k-means [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e], k-modes [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e], hierarchical clustering [\u003cspan additionalcitationids=\"CR15\" citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e] and PAM (Partitioning Around Medoids) paired with Gower distance [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eNo study investigating multimorbidity patterns among prostate cancer hospitalizations has been conducted in the Portuguese context. Therefore, as a first step to understand the implications of multimorbidity clusters on health outcomes such as in-hospital mortality, survival and resource use, we aimed to apply distinct clustering approaches to identify and characterize multimorbidity patterns in a set of inpatient episodes due to prostate cancer using a Portuguese nationwide hospitalization database.\u003c/p\u003e"},{"header":"2. Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1. Data sources and study population\u003c/h2\u003e \u003cp\u003eThis is retrospective observational study using inpatient data from the Portuguese National Hospital Morbidity Database (NHMD). The NHMD contains data on inpatient and outpatient episodes in all mainland public hospitals within the Portuguese National Health Service. The Central Administration for the Health System (ACSS), which is the health authority responsible for the NHMD, provided access to the data. We restricted the analysis to inpatient episodes with a primary diagnosis of prostate cancer (ICD-9-CM code 185 or ICD-10-CM code C61) and with a discharge date between January 1st 2011 and December 31st 2017. Only inpatient episodes labeled in the database as statistically valid were extracted, that is, those with a length of stay of at least 24h, shorter than 24h for patients who died, left against medical advice, or transferred to another hospital.\u003c/p\u003e \u003cp\u003eIn Portugal, the two most common usual treatment paths for hospitalized prostate cancer patients are surgery (radical prostatectomy) and brachytherapy. Radical prostatectomy involves the removal of the entire prostate and the seminal vesicles with or without pelvic lymphadenectomy. It can be used for all risk categories of localized prostate cancer. Alternatively, brachytherapy comprises the insertion of radioactive seeds into the prostate gland, being used for men in low and intermediate risk categories of localized prostate cancer [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. There are other rarer reasons for hospitalization of a prostatic cancer patient that will not be the focus of this research. Therefore, we further restricted the analysis to inpatient episodes reporting radical prostatectomy (ICD-9-CM procedure code 60.5 or ICD-10-PCS codes 0VT00ZZ and 0VT04ZZ) or brachytherapy (ICD-9-CM procedure codes 92.20, 92.27, 92.28 and 92.29 or ICD-10-PCS codes 0VH031Z and chapter DV10), the two main treatment paths for patients admitted with prostate cancer in Portuguese hospitals.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2. Variables\u003c/h2\u003e \u003cp\u003eAs the main goal was to detect patterns of multimorbidity, secondary diagnosis data at episode level was considered to define clusters. These data were transformed into dichotomous variables indicating the presence or absence of Elixhauser comorbidities [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Additionally, dichotomous variables indicating the presence of brachytherapy, surgery and in-hospital death were also considered to form clusters in order to check whether certain multimorbidity patterns may occur along with death or treatment path decision.\u003c/p\u003e \u003cp\u003eMoreover, the identified clusters were further interpreted against other variables such as age, variables reflecting resource use, namely length of stay and relative weight, which measures the episode\u0026rsquo;s cost relative to the average hospitalized national patient, and hospital-related characteristics such as geographic region and hospital group. The latter variable was a category created by ACSS to group hospitals that are similar in terms of case-mix and complexity [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3. Multimorbidity measure\u003c/h2\u003e \u003cp\u003eMultimorbidity was defined as episodes with \u0026ge;\u0026thinsp;2 chronic conditions, including the diagnosis of prostate cancer itself. A total of 30 chronic conditions were identified using the Elixhauser method [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. These conditions were identified based on a predefined list of ICD-9 and ICD-10 codes that were searched across the secondary diagnosis fields in the HDM data. The full list of specific ICD-9 and ICD-10 codes for the 30 Elixhauser conditions can be found in Quan et al. (2005)[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. Cluster analysis was only performed in episodes presenting multimorbidity.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4. Cluster analysis\u003c/h2\u003e \u003cp\u003ePartitioning clustering methods, which are employed to cluster instances based on their level of similarity, can provide a feasible tool for differentiating episodes based on their clinical profile and comorbidity occurrence. The most popular partitioning clustering algorithms is the k-means, but other extensions have been implemented, namely k-modes, and Partitioning Around Medoids (PAM) or CLARA (Clustering Large Applications) algorithms, the latter being an extension of PAM to handle large sample sizes using a sampling approach. [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Since our dataset is only composed of categorical variables, we first tested the k-modes algorithm, which has been widely used to efficiently cluster categorical data [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. K-modes addresses the existing limitation of numeric-only data imposed by the k-means algorithm, enabling clustering of large-size categorical data from real-world databases [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. It adopts a simple dissimilarity measurement adjusted for categorical data denominated total within-cluster simple-matching distance, which compares two instances and computes the number of matching variables relative to the total number of variables.\u003c/p\u003e \u003cp\u003eIn addition to K-modes, PAM and CLARA also extend the k-means clustering algorithm, being less sensitive to outliers and suitable for categorical variables. These approaches find instances in each cluster as a representative object (medoids) and assign all the instances to the clusters they best fit in. PAM starts by finding a medoid, which should be the most centrally located observation in the cluster. Following the determination of the initial set of medoids, one medoid is iteratively replaced by one non-medoid whenever the total distance of the resulting clustering is improved [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. PAM was the preferable method over CLARA considering the study\u0026rsquo;s sample size and computing time.\u003c/p\u003e \u003cp\u003eApart from testing the partitioning methods, hierarchical clustering was also tested. This method is widely used in several fields and produces a hierarchy of clusters represented in a tree-like structure called dendrogram, where it is possible to identify where clusters split and what sub-clusters are subsequently created [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. Hierarchical clustering methods are often regarded as good quality clustering methods, creating clusters of arbitrary shape and are generally robust to noise. There are several approaches to apply hierarchical clustering in the literature. In this study, the agglomerative approach was considered. This approach implements a bottom-up method that starts with all instances forming a separate cluster before being iteratively merged into similarity clusters. In this process, Ward\u0026rsquo;s minimum variance was considered. Ward\u0026rsquo;s method aims at minimizing the total within-cluster variance, in which the pair of clusters with minimum between-cluster distance are merged.\u003c/p\u003e \u003cp\u003eMoreover, Elbow\u0026rsquo;s method and Silhouette analysis [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e] were used to find the optimal number of clusters k. Regarding dissimilarity measurement, k-modes was run considering the total within-cluster simple-matching distance, whereas PAM and Hierarchical clustering were run on a Gower distance matrix calculated from the data, which, unlike other commonly used distance metrics, can handle mixed types of data.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2.5. Statistical analysis\u003c/h2\u003e \u003cp\u003eBaseline characteristics of the included population were summarized by mean, median and interquartile range (IQR) for continuous variables, and frequencies (n) and percentages for categorical variables. Moreover, the Observed to Expected (O/E) prevalence ratios of Elixhauser comorbidities were computed by dividing the prevalence of each comorbidity within each identified cluster by the corresponding prevalence of that condition in the entire analyzed population. Following the approach of Jansana et al. (2021)[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e], a chronic condition was highlighted within a cluster whenever its O/E prevalence ratio was at least 2.\u003c/p\u003e \u003cp\u003eAll statistical analyses were performed using R language and RStudio software. The R package \u0026ldquo;icd\u0026rdquo; was used to identify the Elixhauser conditions in the data according to Quan et al. (2005)[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e] definitions. K-modes clustering required the package \u0026ldquo;klaR\u0026rdquo;, whereas PAM and hierarchical clustering required packages \u0026ldquo;cluster\u0026rdquo; and \u0026ldquo;stats\u0026rdquo;.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cp\u003eThe analyzed dataset included a total of 10394 inpatient episodes, with most of them performing radical prostatectomy (n\u0026thinsp;=\u0026thinsp;8565, 82.0%) (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Nearly half of the included episodes reported multimorbidity (n\u0026thinsp;=\u0026thinsp;6286, 58%), with the most prevalent comorbidities being hypertension (n\u0026thinsp;=\u0026thinsp;4596, 44.0%), diabetes mellitus (n\u0026thinsp;=\u0026thinsp;1413, 14.0%) and obesity (n\u0026thinsp;=\u0026thinsp;826, 7.9%). Additionally, a total of 336 episodes (3.2% of the total) unexpectedly reported metastasis and a very low frequency of in-hospital death (n\u0026thinsp;=\u0026thinsp;31, 0.3%) was observed. The median age was 65 years old (IQR: 60\u0026ndash;69) and the median length of stay was 5 days (IQR:4\u0026ndash;7). Most episodes occurred in the Northern region (n\u0026thinsp;=\u0026thinsp;4215, 41%), followed by Lisbon (n\u0026thinsp;=\u0026thinsp;3050, 30%) and Centre (n\u0026thinsp;=\u0026thinsp;2659, 26%). Hospitals belonging to group E, which is composed of the largest facilities with teaching status, admitted the highest number of episodes (n\u0026thinsp;=\u0026thinsp;3318, n\u0026thinsp;=\u0026thinsp;32%), followed by group F (n\u0026thinsp;=\u0026thinsp;2678, 26%), composed of oncology hospitals, and group C (n\u0026thinsp;=\u0026thinsp;2172, 21%), composed of medium-sized hospitals.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDescriptive Statistics of the analyzed dataset\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCharacteristics\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOverall\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\u003csup\u003e1\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e65 (60\u0026ndash;69)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBrachytherapy\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1830 (18%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRadical Prostatectomy\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e8565 (82%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMetastasis\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e336 (3.2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLength of stay\u003csup\u003e1\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5 (4\u0026ndash;7)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDeath\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e31 (0.3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026zwj;Multimorbidity\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6286 (58%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e\n\u003ch3\u003e1. Median (IQR); 2. Number of episodes (share of the total)\u003c/h3\u003e\n\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e presents the clusters obtained with the different clustering approaches, including selected cluster quality metrics. Each cluster was presented according to the mode of the most prevalent comorbidity or treatment performed. Thus, comorbidity names appearing in the designation of the clusters occurred in at least 50% of the episodes within the clusters. The optimal number of clusters chosen for k-modes was 4, whereas the optimal number of clusters chosen for PAM and hierarchical clustering were 6 and 8, respectively. Clusters obtained with all three methods identified groups with high co-occurrence of hypertension paired with diabetes. However, the K-modes algorithm performed poorly compared to the other approaches as it provided the lowest silhouette scores (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Additionally, K-modes was more unstable, meaning that repeated experiments generated different clusters, apart from producing redundant clusters.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eIdentified multimorbidity clusters among prostate cancer hospitalizations using different clustering approaches.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMethod\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eClusters\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSilhouette index\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eK-modes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCluster 1 (n\u0026thinsp;=\u0026thinsp;3690) - Hypertension (69%) and Surgery (89%)\u003c/p\u003e \u003cp\u003eCluster 2 (n\u0026thinsp;=\u0026thinsp;628) - Hypertension (100%), Diabetes (86%), Surgery (86%)\u003c/p\u003e \u003cp\u003eCluster 3 (n\u0026thinsp;=\u0026thinsp;882) - Hypertension (62%), brachytherapy (58%)\u003c/p\u003e \u003cp\u003eCluster 4 (n\u0026thinsp;=\u0026thinsp;891) - Hypertension (100%), Surgery (88%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.09\u003c/p\u003e \u003cp\u003e0.67\u003c/p\u003e \u003cp\u003e-0.24\u003c/p\u003e \u003cp\u003e0.08\u003c/p\u003e \u003cp\u003e\u003cem\u003eAverage score: \u0026minus;\u0026thinsp;0.006\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePAM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCluster 1 (n\u0026thinsp;=\u0026thinsp;2839) - Hypertension (84%), Surgery (100%)\u003c/p\u003e \u003cp\u003eCluster 2 (n\u0026thinsp;=\u0026thinsp;921) - Hypertension (75%), Diabetes Mellitus (100%), Surgery (100%)\u003c/p\u003e \u003cp\u003eCluster 3 (n\u0026thinsp;=\u0026thinsp;366) - Hypertension (66%), Arrhythmia (100%), Surgery (100%)\u003c/p\u003e \u003cp\u003eCluster 4 (n\u0026thinsp;=\u0026thinsp;620) - Hypertension (72%), Obesity (100%), Surgery (100%)\u003c/p\u003e \u003cp\u003eCluster 5 (n\u0026thinsp;=\u0026thinsp;265) - Chronic pulmonary disease (100%), Surgery (82%)\u003c/p\u003e \u003cp\u003eCluster 6 (n\u0026thinsp;=\u0026thinsp;1080) - Hypertension (78%), Brachytherapy (100%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.49\u003c/p\u003e \u003cp\u003e0.58\u003c/p\u003e \u003cp\u003e0.25\u003c/p\u003e \u003cp\u003e0.30\u003c/p\u003e \u003cp\u003e0.37\u003c/p\u003e \u003cp\u003e0.38\u003c/p\u003e \u003cp\u003e\u003cem\u003eAverage score: 0.45\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHierarchical\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCluster 1 (n\u0026thinsp;=\u0026thinsp;2577) - Hypertension (86%), Surgery (100%)\u003c/p\u003e \u003cp\u003eCluster 2 (n\u0026thinsp;=\u0026thinsp;1008) - Hypertension (76%), Diabetes Mellitus (100%), Surgery (100%)\u003c/p\u003e \u003cp\u003eCluster 3 (n\u0026thinsp;=\u0026thinsp;170) - Depression (100%), Surgery (100%)\u003c/p\u003e \u003cp\u003eCluster 4 (n\u0026thinsp;=\u0026thinsp;319) - Hypertension (64%), Arrhythmia (100%), Surgery (100%)\u003c/p\u003e \u003cp\u003eCluster 5 (n\u0026thinsp;=\u0026thinsp;336) - Chronic pulmonary disease (100%), Surgery (100%)\u003c/p\u003e \u003cp\u003eCluster 6 (n\u0026thinsp;=\u0026thinsp;364) - Hypertension (65%), Obesity (100%), Surgery (100%)\u003c/p\u003e \u003cp\u003eCluster 7 (n\u0026thinsp;=\u0026thinsp;189) - Alcohol abuse (89%), Surgery (100%)\u003c/p\u003e \u003cp\u003eCluster 8 (n\u0026thinsp;=\u0026thinsp;1128) - Hypertension (75%), Brachytherapy (100%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.46\u003c/p\u003e \u003cp\u003e0.44\u003c/p\u003e \u003cp\u003e0.54\u003c/p\u003e \u003cp\u003e0.27\u003c/p\u003e \u003cp\u003e0.44\u003c/p\u003e \u003cp\u003e0.64\u003c/p\u003e \u003cp\u003e0.15\u003c/p\u003e \u003cp\u003e0.33\u003c/p\u003e \u003cp\u003e\u003cem\u003eAverage score: 0.41\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe use of hierarchical and PAM approaches tended to produce more stable and better-quality clusters, with groups presenting clearly defined patterns in terms of comorbidity co-occurrence. Apart from the usual co-occurrence of prostate cancer with hypertension, either among surgical or brachytherapy cases, both methods found three clusters with high co-occurrence of hypertension with diabetes, obesity, and arrhythmia, apart from clusters with high co-occurrence of cancer and chronic pulmonary disease among surgical cases (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Moreover, hierarchical clustering detected clusters of surgical episodes with high co-occurrence of prostate cancer with depression (Cluster 3) or alcohol abuse (Cluster 7).\u003c/p\u003e \u003cp\u003eSince the average silhouette was higher for the clusters obtained with the PAM method, we presented the assessment of the clusters obtained with the PAM method against age, resource use and hospital-related variables (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Only Elixhauser comorbidities with an overall prevalence above 1% are presented in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and highlighted prevalences of comorbidities are related to a O/E ratio higher than 2.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eMultimorbidity clusters obtained with the PAM approach and descriptions based on the prevalence of selected Elixhauser comorbidities and other episode characteristics.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCluster 1\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;2839)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCluster 2\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;921)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCluster 3\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;366)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eCluster 4\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;620)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eCluster 5\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;265)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eCluster 6\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;1080)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eTotal\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;10394)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eNo multimorbidity\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;4303)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eArrhythmia, n(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e366 (100)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e17 (6.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e125 (12)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e508 (4.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHypertension, n(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2377 (84)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e688 (75)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e242 (66)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e445 (72)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e844 (78)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e4596 (44)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChronic pulmonary disease, n(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e151 (5.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e40 (4.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e21 (5.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e40 (6.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e265 (100)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e64 (5.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e581 (5.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiabetes Mellitus, n(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e921 (100)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e53 (14)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e161 (26)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e20 (7.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e258 (24)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1413 (14)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRenal disease, n(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e85 (3.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e27 (2.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e13 (3.6)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e13 (2.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e5 (1.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e29 (2.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e172 (1.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePeptic ulcer disease, n(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e61 (2.1)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e11 (1.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3 (0.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4 (0.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e5 (1.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e18 (1.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e102 (1.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOther Tumors, n(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e105 (3.7)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e15 (1.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e6 (1.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e16 (2.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e7 (2.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e39 (3.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e188 (1.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eObesity, n(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e45 (12)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e620 (100)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e28 (11)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e133 (12)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e826 (7.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAlcohol abuse, n(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e141 (5.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e29 (3.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e13 (3.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e38 (6.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e23 (8.7)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e140 (13)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e384 (3.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDepression, n(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e169 (6.0)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20 (2.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e11 (3.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e22 (3.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e11 (4.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e55 (5.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e288 (2.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMetastasis, n(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e78 (2.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e29 (3.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e9 (2.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e17 (2.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e8 (3.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e59 (5.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e336 (3.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e136 (3.2)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eComorbidity count, median (IQR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1 (1\u0026ndash;1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2 (2\u0026ndash;2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2 (2\u0026ndash;3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2 (2\u0026ndash;3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1 (1\u0026ndash;2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1 (1\u0026ndash;2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1 (0\u0026ndash;2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge, median (IQR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e65 (61\u0026ndash;69)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e65 (62\u0026ndash;69)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e66 (62\u0026ndash;70)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e64 (60\u0026ndash;68)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e65 (60\u0026ndash;68)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e67 (62\u0026ndash;71)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e65 (60\u0026ndash;69)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e64 (59\u0026ndash;68)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLength of stay, median (IQR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6 (4\u0026ndash;7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6 (4\u0026ndash;7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e6 (4\u0026ndash;7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5 (4\u0026ndash;7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e5 (3\u0026ndash;7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e2 (2\u0026ndash;3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e5 (4\u0026ndash;7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e5 (4, 7)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRelative weight, mean (standard deviation)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.13 (\u0026plusmn;\u0026thinsp;0.74)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.19 (\u0026plusmn;\u0026thinsp;0.89)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.30 (\u0026plusmn;\u0026thinsp;0.86)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.17 (\u0026plusmn;\u0026thinsp;0.85)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e2.57 (\u0026plusmn;\u0026thinsp;2.95)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e8.50 (\u0026plusmn;\u0026thinsp;1.25)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e6.36 (\u0026plusmn;\u0026thinsp;1.71)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e5.35 (\u0026plusmn;\u0026thinsp;1.56)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBrachytherapy, n(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e49 (18)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1080 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1830 (18)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e701 (16)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSurgery, n(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2,839 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e921 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e366 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e620 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e216 (82)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e8565 (82)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e3603 (84)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIn-hospital death, n(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1 (\u0026lt;\u0026thinsp;0.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4 (1.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1 (0.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e19 (1.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e31 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e6 (0.1)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHospital group, mode (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGroup E (37)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGroup E (40)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eGroup E (41)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eGroup E (37)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eGroup E (29)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eGroup F (88)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003eGroup E (32)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003eGroup E (32)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGeographic region, mode (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNorth (44)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNorth (40)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNorth (41)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eNorth (37)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eNorth (46)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eCentre (53)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003eNorth (41)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003eNorth (41)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eCluster 6 concentrated nearly all brachytherapy episodes, characterized mostly by the high co-occurrence of hypertension and cancer, although it also presented a O/E ratio of alcohol abuse higher than 2. Also, this cluster presented a substantially higher cost in terms of relative weights when compared to other clusters, despite being composed only of brachytherapy episodes. Accordingly, Cluster 6 presented the highest proportions of in-hospital death and metastasis, suggesting increased severity. Similar to Cluster 6, Cluster 5 also presents a O/E ratio of alcohol abuse higher than 2, with the difference that it comprises mainly surgical episodes presenting high co-occurrence of cancer with chronic pulmonary disease rather than hypertension. Cluster 1 presented an O/E ratio above 2 for peptic ulcer disease, other tumors, and depression, possibly co-occurring with hypertension. Cluster 3, which is only composed of prostate cancer episodes co-occurring with arrhythmia, presented an O/E ratio above 2 for renal disease.\u003c/p\u003e \u003cp\u003eRegarding hospital-related characteristics, the North region was the most representative region across all clusters, except for Cluster 6, which was mostly composed of episodes admitted in hospitals from the Centre region. Furthermore, unlike the observed for other clusters, which were mostly represented by group E hospitals, Cluster 6 concentrated mainly episodes admitted into group F hospitals.\u003c/p\u003e \u003cp\u003eDespite most clusters presenting a very high prevalence of hypertension, none reported an O/E ratio below 2 for this condition due to the high prevalence in the overall included population. Finally, as expected, clusters presenting surgery (radical prostatectomy) reported a higher length of stay. No major differences concerning median age and comorbidity count were observed across the six clusters.\u003c/p\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eTo the best of our knowledge, this is the first study to use clustering analysis to assess multimorbidity patterns in prostate cancer hospitalizations in Portugal, presenting an almost nationwide nature, with a large number of episodes. Different clustering approaches were used, with PAM and hierarchical clustering producing the best results, with similar clusters in terms of main clinical characteristics. These two clustering algorithms were useful and practical approaches to detect multimorbidity patterns, as they clearly differentiated episodes according to the two main treatment paths for hospitalized patients in Portugal (brachytherapy and radical prostatectomy), apart from revealing prevalent co-occurrences of diseases with potential impact on cancer survival and on the treatment itself.\u003c/p\u003e \u003cp\u003eNearly half (58%) of the analyzed prostate cancer episodes reported multimorbidity. According to Rizzuto et al. (2017) [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e], the relationship between multimorbidity and mortality has been extensively studied, however the conclusions have been inconsistent, with some studies reporting a negative association between the presence of multiple comorbidities and survival, whereas others found no association. Considering the negligible number of in-hospital deaths and the pseudonymized nature of our data, which does not allow longitudinal tracking of individuals, we could not find any relationship between multimorbidity and mortality or survival. Furthermore, contrary to previous evidence documented elsewhere in the literature [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e], our results did not find a relationship between the multimorbidity and length of stay. The lower length of stay observed for Cluster 6 (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) was likely to be attributed to the concentration of brachytherapy episodes, whether the remaining clusters were essentially composed of surgical episodes.\u003c/p\u003e \u003cp\u003eSix clusters identified with the PAM method were chosen for a more detailed interpretation (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). The detected clusters presented a similar number of comorbidities, with the highest number consisting of two chronic diseases in addition to prostate cancer (Clusters 2, 3 and 4). Regarding the main patterns observed, Cluster 1 and Cluster 6 were composed exclusively of episodes with a high co-occurence of cancer with hypertension for episodes reporting surgery and brachytherapy, respectively, although Cluster 6 presented the highest burden in terms of costs. Although a clear explanation could not be immediately found, this cluster accounted for the highest prevalence of mortality and metastasis, suggesting that it concentrated more episodes with increased severity and risk of mortality in comparison with the remaining clusters, resulting in higher inpatient costs. Clusters 2, 3 and 4 were composed of surgical episodes reporting a high co-occurrence of hypertension with diabetes mellitus, arrhythmia, and obesity, respectively, in addition to prostate cancer, which was also found using the hierarchical clustering method. In Cluster 5, chronic pulmonary disease co-occurred mostly with prostate cancer alone. The coexistence of several conditions from different medical specialties in the same cluster constitutes a challenging situation for doctors and patients. Although previous literature suggests that integrated care for multimorbid patients can improve health outcomes and efficiency of care, evidence is still limited and may only be feasible in countries with relatively strong and well-resourced health systems [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThese results are also aligned with previous literature, in which prostate cancer diagnosis co-occurred mostly with comorbidities such as hypertension and diabetes, either for localized or advanced patients [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. Moreover, the identified clusters were composed of episodes reporting conditions that may be associated with modifiable risk factors with potential impact on the risk of developing or dying from prostate cancer, such as high fat diet and smoking [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. However, as there is no sufficient evidence on clear indications for prevention beyond early diagnosis to reduce prostate cancer mortality [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e], the focus should be on increasing efforts to manage chronic diseases, namely hypertension and diabetes, during cancer treatment, which includes efforts to control behaviors throughout all phases of prostate cancer survivorship. The identified clusters are clinically relevant in a sense they reveal potential risk groups that should be targeted by different levels of care in order to ensure greater survival and avoid non-planned visits to the health services.\u003c/p\u003e \u003cp\u003eThe geographic distribution of patients found across the clusters corroborates with the National Oncological Registry, which identified regional variations in the incidence of prostate cancer in Portugal in 2018, with the highest incidences in Lisbon, North and Centre regions [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. These regions should thereby be the focus of efforts aimed at implementing oncology referral networks, reinforcing the importance of multidisciplinary teams that can provide integrated care to multimorbid, difficult-to-treat patients. Regarding the treatment path, Cluster 6 suggests an asymmetric distribution of brachytherapy across hospital groups and geographical regions, as most patients were admitted in the Centre region and mostly in oncology hospitals, possibly indicating different treatment practices between regions and hospital types.\u003c/p\u003e \u003cp\u003eIt is important to state the underlying limitations of this study. First, hospital data used for the analyses was obtained from an administrative database, which in turn presents underlying quality issues, mostly related to clinical coding and the degree to which comorbidities are documented and coded [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. Nevertheless, in Portugal, clinical coding is exclusively performed by medical doctors with specific training and verified by internal and external audits. Moreover, as data was de-identified, individual patients could not be identified nor followed. Furthermore, the clustering approaches considered in this study do not consider the temporal changes in the patients’ condition, and episodes are restricted and assumed to be part of a single cluster, rather than the possibility of belonging to more than one cluster simultaneously.\u003c/p\u003e "},{"header":"Conclusions","content":"\u003cp\u003eCluster analysis is a useful approach to detect and characterize different patterns and profiles of prostate cancer hospitalizations in Portugal. Multimorbidity is present in 58% of the admissions for prostate cancer treatment, with these additional conditions potentially accounting for their cause of death, thereby highlighting the importance of addressing the comorbidity status by integrating and coordinating chronic disease care along with cancer care, whenever possible. In particular, the clusters revealed the existence of well-defined and distinct multimorbidity profiles, in which cancer co-occurs mostly with hypertension and other additional chronic disease, such as diabetes, obesity and arrhythmia, potentially incurring a higher treatment burden, requiring thus increased attention during treatment. Future studies should ideally include the assessment of the association between the different identified clusters and survival, recurring to record linkage techniques.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData access was granted by the Central Authority for Health Services, I.P. (ACSS) through a research protocol with the Faculty of Medicine of the University of Porto. \u0026nbsp;Since inpatient data used in this study were previously de-identified by ACSS, there was no need for ethical approval.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data that support the findings of this study are available from the Central Authority for Health Services, I.P. (ACSS). Restrictions apply to the availability of these data, which were used under license for this study. Data are however available from the authors upon reasonable request and with permission of ACSS.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research was funded by Norte2020 (Portugal 2020) within the project Sex Health \u0026amp; Prostate Cancer - Biopsychological Determinants of Sexual Health in Men with Prostate Cancer (NORTE-01-0145-FEDER-000057). It has received Portuguese National Funds through FCT (Portuguese Foundation for Science and Technology) under project UIDB/00760/2020 of the GECAD.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026rsquo; contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA. F. was responsible for the conceptualization; A.F. contributed to the data access; P.C. performed literature review; all authors contributed to the methodology design; P.C. and J.S. conducted data analysis; P.C., J.S., M.L. and F.B. wrote the main manuscript text; J.S. prepared tables 1-3; all authors reviewed the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors would like to thank the Central Authority for Health Services, I.P. (ACSS) for providing access to the data. Sex Health \u0026amp; Prostate Cancer - Biopsychological Determinants of Sexual Health in Men with Prostate Cancer (NORTE-01-0145-FEDER-000057). It has received Portuguese National Funds through FCT (Portuguese Foundation for Science and Technology) under project UIDB/00760/2020 of the GECAD.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSilva Gaspar, S. R., Fernandes, M., Castro, A., Oliveira, T., Santos Dias, J., \u0026amp; Palma Dos Reis, J. Active surveillance protocol in prostate cancer in Portugal. Actas Urol\u0026oacute;gicas Espa\u0026ntilde;olas (English Edition) 46(6), 329\u0026ndash;339 (2022). https://doi.org/10.1016/j.acuroe.2022.01.002\u003c/li\u003e\n\u003cli\u003eGlobal burden of disease study 2019 (GBD 2019) data resources. Healthdata.org., https://ghdx.healthdata.org/gbd-2019, last accessed 2023/04/14\u003c/li\u003e\n\u003cli\u003eJefferson, M., Drake, R. R., Lilly, M., Savage, S. J., Tucker Price, S., \u0026amp; Hughes Halbert, C. Co-morbidities in a Retrospective Cohort of Prostate Cancer Patients. Ethnicity \u0026amp; disease, 30(Suppl 1), 185\u0026ndash;192 (2020). https://doi.org/10.18865/ed.30.S1.185\u003c/li\u003e\n\u003cli\u003ePefoyo, A. J. K., Bronskill, S. E., Gruneir, A., Calzavara, A., Thavorn, K., Petrosyan, Y., Maxwell, C. J., Bai, Y., \u0026amp; Wodchis, W. P. The increasing burden and complexity of multimorbidity. BMC Public Health, 15(1), 415 (2015). https://doi.org/10.1186/s12889-015-1733-2\u003c/li\u003e\n\u003cli\u003eKon\u0026eacute;, A. P., \u0026amp; Scharf, D. Prevalence of multimorbidity in adults with cancer, and associated health service utilization in Ontario, Canada: a population-based retrospective cohort study. BMC cancer, 21(1), 406 (2021). https://doi.org/10.1186/s12885-021-08102-1\u003c/li\u003e\n\u003cli\u003eThe Academy of Medical Sciences. Multimorbidity: A Priority for Global Health Research, https://acmedsci.ac.uk/policy/policy-projects/multimorbidity, last accessed 2023/04/03.\u003c/li\u003e\n\u003cli\u003eThe Academy of Medical Sciences. Multimorbidity: A Priority for Global Health Research. https://acmedsci.ac.uk/policy/policy-projects/multimorbidity, last accessed 2023/08/25.\u003c/li\u003e\n\u003cli\u003eMedical Research Council. Multimorbidity in the UK Population: Understanding Disease Clustering. 2018; https://mrc.ukri.org/funding/browse/multimorbidity/multimorbidity-in-the-uk-population-understanding-disease-clustering/, last accessed 2023/08/25.\u003c/li\u003e\n\u003cli\u003eHaregu T., Oldenburg B., Setswe G., Elliott J. Perspectives, constructs and methods in the measurement of multimorbidity and comorbidity: A critical review. The Internet Journal of Epidemiology, 10(2) (2012). https://doi.org/10.5580/2cc3\u003c/li\u003e\n\u003cli\u003eHarrington, R. L., Qato, D. M., Antoon, J. W., Caskey, R. N., Schumock, G. T., \u0026amp; Lee, T. A. Impact of multimorbidity subgroups on the health care use of early pediatric cancer survivors. Cancer, 126(3), 649\u0026ndash;658 (2020). https://doi.org/10.1002/cncr.32201\u003c/li\u003e\n\u003cli\u003eFillmore, N. R., DuMontier, C., Yildirim, C., La, J., Epstein, M. M., Cheng, D., Cirstea, D., Yellapragada, S., Abel, G. A., Gaziano, J. M., Do, N., Brophy, M., Kim, D. H., Munshi, N. C., \u0026amp; Driver, J. A. Defining multimorbidity and its impact in older United States veterans newly treated for multiple myeloma. Journal of the National Cancer Institute, 113(8), 1084\u0026ndash;1093 (2021). https://doi.org/10.1093/jnci/djab007\u003c/li\u003e\n\u003cli\u003eMcLoone, P., Jani, B. D., Siebert, S., Morton, F. R., Canning, J., Macdonald, S., Mair, F. S., \u0026amp; Nicholl, B. I. Classification of long-term condition patterns in rheumatoid arthritis and associations with adverse health events: a UK Biobank cohort study. Journal of Multimorbidity and Comorbidity, 13, 263355652211486 (2023). https://doi.org/10.1177/26335565221148616\u003c/li\u003e\n\u003cli\u003eBucholc, M., Bradley, D., Bennett, D., Patterson, L., Spiers, R., Gibson, D., Van Woerden, H., \u0026amp; Bjourson, A. J. Identifying pre-existing conditions and multimorbidity patterns associated with in-hospital mortality in patients with COVID-19. Scientific Reports, 12(1), 17313 (2022). https://doi.org/10.1038/s41598-022-20176-w\u003c/li\u003e\n\u003cli\u003eLai, F. T. T., Beeler, P. E., Yip, B. H. K., Cheetham, M., Chau, P. Y. K., Chung, R. Y., Wong, E. L. Y., Yeoh, E.-K., Battegay, E., \u0026amp; Wong, S. Y. S. Comparing multimorbidity patterns among discharged middle-aged and older inpatients between Hong Kong and Zurich: A hierarchical agglomerative clustering analysis of routine hospital records. Frontiers in Medicine, 8, 651925 (2021). https://doi.org/10.3389/fmed.2021.651925\u003c/li\u003e\n\u003cli\u003eCezard, G., Sullivan, F., \u0026amp; Keenan, K. Understanding multimorbidity trajectories in Scotland using sequence analysis. Scientific Reports, 12(1), 16485 (2022). https://doi.org/10.1038/s41598-022-20546-4\u003c/li\u003e\n\u003cli\u003eJansana, A., Poblador-Plou, B., Gimeno-Miguel, A., Lanzuela, M., Prados-Torres, A., Domingo, L., Comas, M., Sanz-Cuesta, T., Del Cura-Gonzalez, I., Iba\u0026ntilde;ez, B., Abizanda, M., Duarte-Salles, T., Padilla-Ruiz, M., Redondo, M., Castells, X., Sala, M., \u0026amp; SURBCAN Group. Multimorbidity clusters among long-term breast cancer survivors in Spain: Results of the SURBCAN study. International Journal of Cancer. Journal International Du Cancer, 149(10), 1755\u0026ndash;1767 (2021). https://doi.org/10.1002/ijc.33736\u003c/li\u003e\n\u003cli\u003eRobertson, L., Vieira, R., Butler, J., Johnston, M., Sawhney, S., \u0026amp; Black, C. Identifying multimorbidity clusters in an unselected population of hospitalised patients. Scientific Reports, 12(1), 5134 (2022). https://doi.org/10.1038/s41598-022-08690-3\u003c/li\u003e\n\u003cli\u003eMottet N., van den Bergh R.C.N., Briers E., Van den Broeck T., Cumberbatch M.G., De Santis M., Fanti S., Fossati N., Gandaglia G., Gillessen S., Grivas N., Grummet J., Henry A.M., van der Kwast T.H., Lam T.B., Lardas M., Liew M., Mason M.D., Moris L., Oprea-Lager D.E., van der Poel H.G., Rouvi\u0026egrave;re O., Schoots I.G., Tilki D., Wiegel T., Willemse P.M., Cornford P. EAU-EANM-ESTRO-ESUR-SIOG Guidelines on Prostate Cancer-2020 Update. Part 1: Screening, Diagnosis, and Local Treatment with Curative Intent. European urology, 79(2), 243\u0026ndash;262 (2021). https://doi.org/10.1016/j.eururo.2020.09.042\u003c/li\u003e\n\u003cli\u003eElixhauser, A., Steiner, C., Harris, D. R., \u0026amp; Coffey, R. M. Comorbidity measures for use with administrative data. Medical Care, 36(1), 8\u0026ndash;27 (1998). https://doi.org/10.1097/00005650-199801000-00004\u003c/li\u003e\n\u003cli\u003eSouza, J., Caballero, I., Vasco Santos, J., Lobo, M., Pinto, A., Viana, J., S\u0026aacute;ez, C., Lopes, F., \u0026amp; Freitas, A. Multisource and temporal variability in Portuguese hospital administrative datasets: Data quality implications. Journal of Biomedical Informatics, 136(104242), 104242 (2022). https://doi.org/10.1016/j.jbi.2022.104242\u003c/li\u003e\n\u003cli\u003eQuan, H., Sundararajan, V., Halfon, P., Fong, A., Burnand, B., Luthi, J. C., Saunders, L. D., Beck, C. A., Feasby, T. E., \u0026amp; Ghali, W. A. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Medical care, 43(11), 1130\u0026ndash;1139 (2005). https://doi.org/10.1097/01.mlr.0000182534.19832.83\u003c/li\u003e\n\u003cli\u003eHao, M., Li, Y., \u0026amp; Yamamoto, T. Public preferences and willingness to pay for shared autonomous vehicles services in Nagoya, Japan. Smart Cities, 2(2), 230\u0026ndash;244 (2019). https://doi.org/10.3390/smartcities2020015\u003c/li\u003e\n\u003cli\u003eZhou, H., Zhang, Y., \u0026amp; Liu, Y. A Global-Relationship Dissimilarity Measure for the k-Modes Clustering Algorithm. Computational intelligence and neuroscience, 2017, 3691316 (2017). https://doi.org/10.1155/2017/3691316\u003c/li\u003e\n\u003cli\u003eBai, L., \u0026amp; Liang, J. The k-modes type clustering plus between-cluster information for categorical data. Neurocomputing, 133, 111\u0026ndash;121 (2014). https://doi.org/10.1016/j.neucom.2013.11.024\u003c/li\u003e\n\u003cli\u003eJohnson, S. C. Hierarchical clustering schemes. Psychometrika, 32(3), 241\u0026ndash;254 (1967). https://doi.org/10.1007/bf02289588\u003c/li\u003e\n\u003cli\u003eRousseeuw, P. J. Silhouettes: A graphical aid to the interpretation and validation of cluster analysis. Journal of Computational and Applied Mathematics, 20, 53\u0026ndash;65 (1987). https://doi.org/10.1016/0377-0427(87)90125-7\u003c/li\u003e\n\u003cli\u003eRizzuto, D., Melis, R. J. F., Angleman, S., Qiu, C., \u0026amp; Marengoni, A. Effect of chronic diseases and multimorbidity on survival and functioning in elderly adults. Journal of the American Geriatrics Society, 65(5), 1056\u0026ndash;1060 (2017). https://doi.org/10.1111/jgs.14868\u003c/li\u003e\n\u003cli\u003eBayer-Oglesby, L., Zumbrunn, A., Bachmann, N., \u0026amp; on behalf of the SIHOS Team. Social inequalities, length of hospital stay for chronic conditions and the mediating role of comorbidity and discharge destination: A multilevel analysis of hospital administrative data linked to the population census in Switzerland. PloS One, 17(8), e0272265 (2022). https://doi.org/10.1371/journal.pone.0272265\u003c/li\u003e\n\u003cli\u003eNavickas, R., Petric, V. K., Feigl, A. B., \u0026amp; Seychell, M. Multimorbidity: What do we know? What should we do?. Journal of comorbidity, 6(1), 4\u0026ndash;11 (2016). https://doi.org/10.15256/joc.2016.6.72\u003c/li\u003e\n\u003cli\u003eBenzo, R., Moreno, P. I., Fox, R. S., Silvera, C. A., Walsh, E. A., Yanez, B., Balise, R. R., Oswald, L. B., \u0026amp; Penedo, F. J. Comorbidity burden and health-related quality of life in men with advanced prostate cancer. Research square, rs.3.rs-2572781 (2023). https://doi.org/10.21203/rs.3.rs-2572781/v1\u003c/li\u003e\n\u003cli\u003eAdjakly, M., Ngollo, M., Dagdemir, A., Judes, G., Pajon, A., Karsli-Ceppioglu, S., Penault-Llorca, F., Boiteux, J. P., Bignon, Y. J., Guy, L., \u0026amp; Bernard-Gallon, D. Prostate cancer: The main risk and protective factors-Epigenetic modifications. Annales d\u0026apos;endocrinologie, 76(1), 25\u0026ndash;41 (2015). https://doi.org/10.1016/j.ando.2014.09.001\u003c/li\u003e\n\u003cli\u003eBrookman-May, S. D., Campi, R., Henr\u0026iacute;quez, J. D. S., Klatte, T., Langenhuijsen, J. F., Brausi, M., Linares-Espin\u0026oacute;s, E., Volpe, A., Marszalek, M., Akdogan, B., Roll, C., Stief, C. G., Rodriguez-Faba, O., \u0026amp; Minervini, A. Latest Evidence on the Impact of Smoking, Sports, and Sexual Activity as Modifiable Lifestyle Risk Factors for Prostate Cancer Incidence, Recurrence, and Progression: A Systematic Review of the Literature by the European Association of Urology Section of Oncological Urology (ESOU). European urology focus, 5(5), 756\u0026ndash;787 (2019). https://doi.org/10.1016/j.euf.2018.02.007\u003c/li\u003e\n\u003cli\u003eGandaglia, G., Leni, R., Bray, F., Fleshner, N., Freedland, S. J., Kibel, A., Stattin, P., Van Poppel, H., \u0026amp; La Vecchia, C. Epidemiology and Prevention of Prostate Cancer. European urology oncology, 4(6), 877\u0026ndash;892 (2021). https://doi.org/10.1016/j.euo.2021.09.006\u003c/li\u003e\n\u003cli\u003eMinistry of Health (2018). Registo oncol\u0026oacute;gico nacional de todos os tumores na popula\u0026ccedil;\u0026atilde;o residente de Portugal. Retrieved from: https://ron.min-saude.pt/media/2196/2021-0518_publica%C3%A7%C3%A3o-ron_2018.pdf, last accessed 23/08/18\u003c/li\u003e\n\u003cli\u003eSouza, J., Santos, J. V., Canedo, V. B., Betanzos, A., Alves, D., \u0026amp; Freitas, A. (2020). Importance of coding co-morbidities for APR-DRG assignment: Focus on cardiovascular and respiratory diseases. Health information management: journal of the Health Information Management Association of Australia, 49(1), 47\u0026ndash;57 (2020). https://doi.org/10.1177/1833358319840575\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Clustering, Inpatient, Prostate Cancer, Machine Learning, Multimorbidity","lastPublishedDoi":"10.21203/rs.3.rs-4247648/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4247648/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eMultimorbidity is a common condition among cancer patients, resulting in increased complexity of care and risk of negative outcomes. This study aims to use clustering analysis to identify and characterize multimorbidity patterns among hospitalized prostate cancer patients in Portugal. This is a retrospective observational study using inpatient data from the Portuguese National Hospital Morbidity Database. Data on hospital admissions with a diagnosis of prostate cancer occurring in all public hospitals in mainland Portugal during 2011\u0026ndash;2017 were considered. Partitioning clustering algorithms, namely K-modes, PAM (Partitioning Around Medoids), and hierarchical clustering, were used to identify multimorbidity clusters. Results obtained from the different clustering approaches were compared and assessed in terms of clinical relevance. A total of 10394 inpatient episodes were analyzed, with 6091 (58%) reporting multimorbidity. Similar clusters were obtained through the different partitioning approaches, with PAM presenting a higher stability and the best quality results in terms of average silhouette. The analysis of the 6 clusters obtained with PAM reveals groups with a pattern of hypertension co-occurring with diabetes, obesity, and arrhythmia, in addition to cancer itself. In this study, the validity of cluster analysis as an exploratory method for identifying clusters of multimorbid conditions among prostate cancer patients in Portugal was demonstrated, identifying relevant patterns of disease co-occurrence, with potential impact on treatment decisions and outcomes. The identified clusters revealed conditions that typically co-occur with prostate of cancer and that can be controlled throughout all phases of cancer survivorship by means of healthier behaviors aligned with integrated and coordinated care.\u003c/p\u003e","manuscriptTitle":"Multimorbidity patterns among patients hospitalized with prostate cancer in Portugal: a cluster analysis approach","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-04-16 12:58:04","doi":"10.21203/rs.3.rs-4247648/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"c6a849c4-a51c-4f94-8374-2fafa35df734","owner":[],"postedDate":"April 16th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-10-13T23:08:12+00:00","versionOfRecord":[],"versionCreatedAt":"2024-04-16 12:58:04","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4247648","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4247648","identity":"rs-4247648","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}