Hospitalization rates for complications due to systemic therapy in the United States

The aim of this study was to estimate the trends and burdens associated with systemic therapy-related hospitalizations, using nationally representative data. National Inpatient Sample data from 2005 to 2016 was used to identify systemic therapy-related complications using ICD-9 and ICD-10 external causes-of-injury codes. The primary outcome was hospitalization rates, while secondary outcomes were cost and in-hospital mortality. Overall, there were 443,222,223 hospitalizations during the study period, of which 2,419,722 were due to complications of systemic therapy. The average annual percentage change of these hospitalizations was 8.1%, compared to − 0.5% for general hospitalizations. The three most common causes for hospitalization were anemia (12.8%), neutropenia (10.8%), and sepsis (7.8%). Hospitalization rates had the highest relative increases for sepsis (1.9-fold) and acute kidney injury (1.6-fold), and the highest relative decrease for dehydration (0.21-fold) and fever of unknown origin (0.35-fold). Complications with the highest total charges were anemia ($4.6 billion), neutropenia ($3.0 billion), and sepsis ($2.5 billion). The leading causes of in-hospital mortality associated with systemic therapy were sepsis (15.8%), pneumonia (7.6%), and acute kidney injury (7.0%). Promoting initiatives such as rule OP-35, improving access to and providing coordinated care, developing systems leading to early identification and management of symptoms, and expanding urgent care access, can decrease these hospitalizations and the burden they carry on the healthcare system.

. NIS has enabled researchers and policymakers to estimate parameters such as healthcare utilization and outcomes, hospitalization cost, access and quality of care, and total healthcare spending at the national level, which are useful for healthcare policy decisions. Each year, the NIS collects data from more than 7 million hospitalizations in the United Sates, which corresponds to 35 million weighted hospitalizations. To increase its ability in capturing national estimates, the NIS redesigned its data collection methods in 2012. The NIS currently collects and stores a sample of discharge records from all hospitals participating in HCUP, rather than all discharge records collected from a sample of hospitals (as was its practice prior to 2012). Hospitalization data within the NIS contains one primary and as many as 29 secondary diagnoses, in addition to other variables, such as demographics, hospital data, clinical procedures, length of stay, disposition status, and total charges. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes and Tenth Revision (ICD-10-CM) codes were used for reporting primary and secondary diagnoses in this study.
Study design. The current study was a retrospective analysis of NIS data collected from 2005 to 2016. Cancer patients were identified through Clinical Classifications Software (CCS) codes 11-45 indicating neoplasms as described previously 14 . Cancers were categorized into solid tumors and hematologic malignancies ( Table 1).
Complications of systemic therapy were identified using ICD-9-CM and ICD-10-CM external causes-of-injury codes (E-codes) (Supplement File). The first listed non-cancer diagnosis identified using the ICD-9-CM or ICD-10-CM diagnosis code as the primary reason for hospitalization (eTable S1). The procedures used for selecting the patient cohort are presented in the CONSORT diagram (eFig. S1). Other variables included demographics (age, sex, race), socioeconomic factors (median household income by zip code, insurance type), and hospital characteristics (region, bed size, and teaching status). Primary outcome variables included hospitalization rate and total cost; secondary outcome variables included hospital length of stay and in-hospital mortality for complications. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for reporting our findings. The methods for this study were conducted in accordance with relevant www.nature.com/scientificreports/ guidelines and regulations. We followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines for reporting our findings. The study was reviewed by the Miami Cancer Institute's Institutional Review Board (IRB), which exempted the study from IRB approval. The IRB also waived the requirement for informed consent because it used previously collected de-identified data stored in NIS.
Statistical analysis. Statistical analyses were performed using SAS (version 9.4, SAS Institute, Cary, North Carolina), which accounts for the complex survey design and clustering. The guidelines for using NIS data developed by Khera and Krumholz were used to ensure appropriate procedures for the study 15 . As already mentioned, the NIS was redesigned in 2012 to improve national estimates. To account for these changes, we used modified discharge weights for the years 2005-2011 16 . Descriptive statistics were used to understand temporal factors, types of cancers, demographics, socioeconomic factors, and hospital characteristics. Hospitalization rates were calculated by dividing the total number of patients hospitalized due to complications of systemic therapy by the total number of hospitalizations. Costs of individual inpatient-stays were calculated by multiplying total hospital charges by cost-to-charge ratios. The costs for each year were adjusted according to 2016 inflation levels, based on the US Consumer Price Index. Hospital length of stay was estimated by subtracting the admission date from the discharge date. In-hospital mortality was calculated by dividing the number of patients admitted for complications of systemic therapy and those who died in the hospital by the total number of patients hospitalized due to complications of systemic therapy. All analyses included weighted frequencies for calculating national estimates. Statistical significance was set at P < 0.05.

Discussion
To the best of our knowledge, this is the first study that reports on hospitalizations for complications of systemic therapy, using a nationally representative dataset. In this study, hospitalizations for complications of systemic therapy increased significantly, in contrast to a decrease in general hospitalizations for all causes during the same study period. A significant majority of these hospitalizations consisted of older adults greater than 60 years of age on Medicare benefits and admitted to urban teaching centers. Complications associated with systemic therapy that resulted in the highest number of hospitalizations were anemia, neutropenia, and sepsis.  www.nature.com/scientificreports/ The observed increase in hospitalizations for complications of systemic therapy could be explained by an increase in the number of elderly patients who are becoming eligible for systemic therapy 1 . As newer targeted treatments have significantly improved overall survival and progression-free survival rates in the overall population, these benefits have also been translated to elderly patients as well 17,18 . Nevertheless, these new agents have also increased the number of adverse events because of side effects, in addition to those already resulting from existing treatment regimens 19 . These factors could be responsible for increasing the volume of hospitalization for managing these conditions. The majority of the patients were admitted in large urban teaching hospitals as cancer management requires intensive treatment and follow-up most widely available in higher-tiered hospitals 20 .
Among patients receiving systemic therapy for both solid tumors and hematologic malignancies, anemia, neutropenia, and sepsis were the most common complications requiring hospitalization. This could be attributed to profound myelosuppression and immunosuppression associated with chemotherapy and bone marrow transplantation. In addition, bone marrow transplant recipients are more likely to experience readmissions for infections and graft failures in the initial month after discharge from a primary hospitalization 21 . Nausea and vomiting were common among patients receiving systemic therapy for bone and connective tissue cancers. These cancers are treated with consecutive-day regimens of multiple anti-cancer drugs, resulting in higher levels of nausea and vomiting prominent in cisplatin-containing regimens 22 . Dehydration was common among colon and head and neck cancers. This could be due to complications such as mucositis, common in these patients, limiting their ability to consume food and beverages, as well as destruction of the lining of the gastrointestinal tract, limiting the absorption of fluids and electrolytes 23 . Acute kidney injury was common among Non-Hodgkin lymphoma and leukemia patients. Kidney injury in these conditions could be caused by the malignancy itself or as a side effect of systemic therapy 24 . Congestive heart failure was common among non-Hodgkin lymphoma and breast cancer patients. This finding could be due to the adverse effects of medications such as trastuzumab and anthracycline, commonly recommended for these cancers 25 .
In our study, sepsis, pneumonia, and aspiration pneumonitis were associated with prolonged length of stay and mortality among those receiving systemic therapy. This is not surprising as cancer treatments cause significant immunosuppression and disruption of mucosal barriers. Some studies have reported a significant increase in the length of stay and higher mortality rates when cancer was complicated by severe sepsis 26,27 . Pneumonia and aspiration pneumonitis are usually severe in leukemia and lung cancer patients and hence associated with greater length of stay and mortality 28,29 .
The complications that were associated with the highest total charges were anemia, sepsis, neutropenia, and pneumonia. Though sepsis had the highest median cost, total charges were highest for anemia because of the large volume of patients admitted for this condition. Previous studies evaluating the financial burdens of cancer hospitalization have shown similar findings. For example, in a large-scale study among breast cancer patients, anemia and neutropenia had the highest expenditures 30 . Similarly, in a retrospective study among 412,005 cancer patients of all types, sepsis was among the 3 most costly complications 29 .
Hospitalizations for complications of systemic therapy involve significant financial burdens in addition to poor quality of life 29 . Efforts to lower these expenditures could enormously decrease healthcare costs. An important step in this direction would be to accurately identify unplanned hospitalizations in cancer care. Rule OP-35, proposed by the Centers for Medicare and Medicaid Services (CMS), offers hope in this direction 31 . This rule proposes to identify patients who have received systemic therapy in outpatient settings and subsequently hospitalized for 10 major complications such as anemia, neutropenia, sepsis, pneumonia, nausea, vomiting, dehydration, pain, diarrhea and fever, within 30 days. This process estimates the care provided to cancer patients, which is   www.nature.com/scientificreports/ subsequently analyzed and improved with a focus on decreasing the number of unplanned hospitalizations. In addition to these measures, other strategies such as improving access to and providing coordinated care, developing standardized procedures for prompt identification and management of symptoms, and implementing urgent cancer care strategies, can significantly decrease not only hospitalization rates but also the duration of stay and the severity of these complications 32 .
Limitations. In our study, we used E codes to identify patients who experienced hospitalizations for complications of systemic therapy. However, since the NIS is primarily an administrative database, E codes could be underrepresented because they are not obligatory for reimbursements. This may have resulted in an underestimation of the volume of hospitalizations for these conditions. Furthermore, the NIS does not have data on tumor staging or anti-cancer medications. Therefore, we could not stratify the complications based on disease severity or treatment regimen, thus limiting our findings. In addition, NIS erases all personal identifiers for data confidentiality. Thus, the same patient readmitted multiple times would have been considered as an independent new admission, leading to an overestimation of hospitalization rates. Despite these limitations, we decided to use NIS because data on hospitalization for complications of cancer treatment is sparse. In addition, we could accurately calculate national estimates due to the large sample size provided by the NIS, which is the largest, all payer in-patient database in the United States.

Conclusions
The management of complications of cancer treatment pose a significant burden to hospitals. From 2005 to 2016, hospitalization rates for complications of systemic therapy increased by an annual rate of 8.1%, which is concerning, when compared to the fact that the overall hospitalization rate decreased by an annual rate of 0.5% during the same period. The most common complications of systemic therapy that required hospitalization were anemia, neutropenia, and sepsis. Among all complications, sepsis was associated with greatest length of stay, highest mortality, and highest expenditures. Improved strategies for identifying and managing these complications in the outpatient and emergency settings could significantly decrease these hospitalizations, the burden on the healthcare system, and subsequently improve the quality of life of cancer patients.