Preventable and mitigable adverse events in cancer care: Measuring risk and harm across the continuum

Abstract

BACKGROUND:

Patient safety is a critical concern in clinical oncology, but our ability to measure adverse events (AEs) across cancer care is limited by a narrow focus on treatment-related toxicities. Our objective was to assess the nature and extent of AEs among cancer patients across inpatient and outpatient settings.

METHODS:

Retrospective cohort study of 400 adult patients with breast (n=128), colorectal (n=136), or lung (n=136) cancer treated at a comprehensive cancer center in 2012 and followed for 1 year. Trained reviewers screened both medical records using an oncology trigger tool and safety reporting databases to identify candidate AEs, or injuries due to medical care. Physicians determined harm severity and likelihood of preventability and harm mitigation.

RESULTS:

The 400 patient sample represented 133,358 follow-up days. We identified 304 AEs, for an overall rate of 2.3 events per 1000 patient days (91.2 per 1000 inpatient days; 0.9 per 1000 outpatient days). Thirty-four percent of patients had ≥1 AE (95% CI: 29-39%) and 16% of patients had ≥1 preventable or mitigable AE (95% CI: 13-20%). The AE rate among patients with breast cancer was lower than that for patients with colorectal or lung cancers (p≤0.001). The preventable or mitigable AE rate was 0.9 per 1000 patient days. Six percent of AEs and 4% of preventable AEs resulted in serious harm. Examples included lymphedema, abscess, and renal failure.

CONCLUSION:

We identified a heavy burden of AEs, including preventable or mitigable events. Future research should examine risk factors and improvement strategies to reduce their burden.

Condensed abstract:

The purpose of this study was to assess the nature and extent of adverse events among cancer patients across inpatient and outpatient settings. A heavy burden of adverse events was identified, highlighting opportunities for oncology clinicians to think more broadly about oncology-related harm.

INTRODUCTION

Patient safety is a key focus in clinical oncology. Successful therapy requires a thoughtful balance of treatment-related toxicities and long-term recurrence-free survival. Management of toxicities is a core competence of oncology practitioners, and advances in therapeutics have relied heavily on innovations in symptom management, dose titration, and supportive therapies.

Given an intense focus on treatment-related toxicities, it is surprising that oncology lags other areas of medicine in its understanding of the nature and extent of medical errors and injuries. The ability to measure the frequency, spectrum, and preventability of adverse events (AEs) among cancer patients and across settings is limited to a few small and dated studies^{1, 2} This is an area of unmet need, given the potential severity of illness among cancer patients, its myriad manifestations and complications, and the toxicities of therapy. A deeper understanding of the nature and extent of AEs in cancer care may offer insights that can inform interventions to improve patient outcomes and reduce suffering.

The few studies of patient safety in oncology suggest that AEs are common and harmful. In one study of patients receiving outpatient chemotherapy at a single cancer center in 2000, the medication order error rate was 3%, and 2% of errors had potential for harm.¹ In a more recent study of 4 US outpatient oncology clinics, the rate of potentially injurious errors was 5 per 1000 medication orders in adults and 10 per 1000 in children.² These rates of errors and AEs in ambulatory oncology are low compared to the rate of adverse drug events in studies of hospitalized (5-10%)^3–11 and ambulatory (25%)^{1, 12} general medicine patients, raising the possibility of systematic under-estimation of the problem’s magnitude.

While clinicians can recognize risk of harm for individual patients, empirical evidence is limited regarding the rate at which mistakes are made, the types of mistakes, their impact on patients, and opportunities for improvement. Therefore, we conducted a study to assess the nature and extent of harm in the form of AEs, or unintended harm associated with medical care, among cancer patients longitudinally across inpatient and outpatient care at a comprehensive cancer center. We focused on preventable events and events where the severity or duration of harm could have been mitigated.

METHODS

Study design

We conducted a retrospective cohort study of 400 patients at a single comprehensive cancer center to identify AEs (i.e., injuries associated with medical care) and potential preventable or mitigable AEs during the course of cancer care in both inpatient and outpatient settings.

Setting and participants

We conducted this study at Memorial Sloan Kettering Cancer Center (MSK), a New York-based NCI-designated comprehensive cancer center with a full range of services for adult and pediatric cancer patients. We studied a cohort of 400 patients aged 18 years or older diagnosed with breast, colorectal, or lung cancer who began their first cancer-directed treatment between January 1, 2012 and December 31, 2012. We followed each patient for up to one year or death, whichever came first. We used stratified random sampling to ensure a diverse and representative cohort. For patients with breast cancer we stratified by stage and chemotherapy use, for colorectal cancer, stage and colon/rectal cancer, and for lung cancer, stage and non-small cell lung cancer/small cell lung cancer (Online-only supplemental material 1).

Data sources

MSK uses a homegrown Healthcare Information System (HIS) for storage and retrieval of patient information across settings within the institution. HIS is an integrated clinical application that includes laboratory, pathology, and radiology test results, surgical procedures, chemotherapy and pharmacy profiles, treatment pathways and guidelines, an order management system, electronic order entry and electronic signature capabilities, the electronic medical record, and patient characteristics. Inpatient and outpatients records are included in the HIS.

Information about AEs that are reported by clinicians is maintained at MSK in the Surgical Secondary Events and RL6:RISQ (RL Solutions, Toronto, ON) systems. These systems are separate from the HIS. The Surgical Secondary Events database includes AE reports that are submitted by clinicians for surgical cases. RL6: RISQ includes reports of AEs, errors, and ‘near misses’ that are submitted voluntarily by any front-line staff member.

Study outcomes and variables

AE was defined as unintended harm to the patient by act of commission or omission rather than by the underlying disease or condition of the patient.^{6, 13, 14} Adverse drug events (ADEs) are included as a sub-set of AEs.

An event was deemed ‘preventable’ if the AE resulted from clinical care that was inconsistent with standard oncology practice, or a treatment-related complication that should have been anticipated. Events deemed not likely to be preventable were further evaluated for their likelihood of being ‘mitigable.’ An event was deemed ‘mitigable’ if the severity or the duration of harm could have been lessened had clinicians acted promptly and appropriately.¹⁵ Likelihood of preventability and mitigation were classified as: definitely, probably, probably not, definitely not, or unable to determine. We defined severity of harm associated with the AE based on the National Coordinating Council for Medication Error Reporting and Prevention Index.¹⁶

Patient demographic and administrative variables included: age, sex, race, ethnicity, non-English primary language, marital status, insurance status, comorbid conditions, disease stage, and cancer treatments received (i.e., chemotherapy, radiation therapy, surgery). These variables were obtained electronically from the HIS except comorbid conditions, which were manually abstracted by physician reviewers from patient intake forms.

Data abstraction

To identify candidate AEs, experienced oncology nurses reviewed all available medical records for each patient for one year following the first cancer-directed therapy at MSK. Records included information about in- and outpatient care at MSK, tests and treatments received at MSK, and any information about non-MSK care that was recorded or archived in the HIS. Nurses reviewed records to identify candidate AEs.

To facilitate the nurses’ record review, the study team developed an oncology-based screening tool.^{17, 18} This tool included a list of ‘triggers’ or signals that would lead reviewers to a focused review of the record for the occurrence of AEs. Details of the development of the tool have been previously published.^17–20 If a ‘trigger’ or signal was identified, the reviewer performed a more detailed record abstraction to assess whether a candidate AE occurred. For candidate AEs, nurse reviewers recorded case details, patient information, date of event, severity of harm, and related outcomes. A maximum of one hour was allotted per case. Investigators trained nurse reviewers in chart abstraction, use of the screening tool, and standardized documentation. Training included a review of a sample of cases to ensure consistency between reviewers.

We also instructed nurses to include a review of existing AE records from the Surgical Secondary Events and RL6:RISQ systems. During the initial medical record review, nurse reviewers were blinded to the existing AE reports. Upon completion of each review, the nurses were instructed to open a separate document with any AE reports. They were instructed to review the event and assess whether it met the criteria for inclusion in this study, note if they had already identified it, and if not, review the record for evidence of the event.

Event classification

After the nurse reviewers completed the initial AE identification using the trigger tool as a guide, they presented each candidate AE to a pair of physician reviewers. The physician reviewers were instructed by the study team to independently classify whether the event represented an AE per the study definition, the severity of harm associated with the AE, the likelihood that the event could have been prevented, and, if not preventable, the likelihood that harm that resulted from the event could have been mitigated. Physicians had the opportunity to ask questions about the case, and nurses were encouraged to access the patient’s medical record as needed. Physicians discussed their responses and reached a consensus.

We calculated the percent agreement between physician pairs with regards to whether or not the case was an AE (98%), level of harm severity (A-D compared to E-I) (79%), likelihood of preventability (definitely or probably preventable; definitely or probably not preventable; or unable to determine) (78%), and likelihood of harm mitigation (definitely or probably mitigable; definitely or probably not mitigable; or unable to determine) (83%).

Analyses

We examined the total number of AEs and preventable or mitigable AEs, by cancer type. We estimated the average time from the first MSK visit to the AEs. We examined AE harm severity, by cancer type and preventability. We also calculated the rate of AEs and preventable or mitigable AEs per 1000 total patient days and per 100 patients, and by setting of care: AEs and preventable or mitigable AEs per 1000 inpatient days, per 1000 outpatient days, and per 100 hospital admissions.

We calculated the proportion of patients with at least one AE and at least one preventable or mitigable AE, by cancer type and stage of disease. We obtained the proportion of patients with multiple AEs and duration of time between AEs. We assessed the extent of overlap in AEs identified between data sources: medical records versus existing institutional databases (Surgical Secondary Events and RL6:RISQ). The study was considered Exempt research by the MSK Institutional Review Board. All analyses were conducted using Microsoft Excel and SAS Software (SAS Institute).

RESULTS

Patient characteristics

We reviewed medical records for patients with breast (n=128), colorectal (n=136), and lung (n=136) cancers, which represented 133,358 total days of follow-up. Demographic, administrative and clinical characteristics of the patient cohort are included in Table 1. Patients’ mean age at the start of the study was 61 years, 32% were male, 19% were non-white, and 6% were Hispanic/Latino.

Table 1.

Cohort characteristics

Characteristics	Total %	Breast Cancer %	Colorectal Cancer %	Lung Cancer %
Total (n)	400	128	136	136
Mean age at start of study, SD	61 yrs, 13 yrs	55 yrs, 13 yrs	60 yrs, 13 yrs	67 yrs, 11 yrs
Male	32%	NA	51%	42%
Non-white race	19%	23%	20%	15%
Hispanic/Latino	6%	5%	8%	6%
Non-English primary language	7%	6%	8%	7%
Married (vs other)	66%	62%	68%	68%
Insurance status
Commercial	55%	70%	56%	40%
Medicare	43%	27%	42%	57%
Medicaid	2%	2%	2%	2%
Self-pay	1%	1%	0%	1%
Mean no. of comorbidities, SD1	1.2, 1.3	0.9, 1.0	1.1, 1.3	1.6, 1.4
Early stage of disease [In-situ - III] (vs advanced)	57%	74%	48%	50%
Cancer-directed treatment
Surgery	74%	78%	82%	62%
Chemotherapy	66%	69%	61%	68%
IV chemotherapy	86%	72%	98%	90%
Oral chemotherapy	37%	49%	30%	33%
Radiation therapy	33%	42%	15%	42%

Percents may not add to 100% due to rounding.

¹Comorbid conditions included: myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, connective tissue disease, ulcer disease, mild liver disease, diabetes, hemiplegia, moderate or severe renal disease, diabetes with end organ damage, moderate or severe liver disease, AIDS, hyperlipidemia, hypertension, and coronary artery disease.

Overall rates and harm

We identified 304 unique AEs, for an overall rate of 2.3 events per 1000 patient days. Ninety-seven AEs (33% of total AEs) were deemed definitely or probably preventable. Among the 147 events deemed non-preventable, harm could have definitely or probably been mitigated in 18 (12%). The overall rate of preventable AEs was 0.73 per 1000 patient days and the rate of mitigable AEs was 0.13 per 1000 patient days. The largest proportion of preventable or mitigable AEs was in lung cancer (47%), followed by colorectal cancer (36%), and breast cancer (20%) (Table 2 and Online-only supplemental material 2).

Table 2.

Adverse events (AEs) over a 1 year time period, by cancer type

	Total	Breast Cancer	Colorectal Cancer	Lung Cancer
Total AEs
No. of AEs identified	304	46	135	123
AEs / 100 patients	76	36	99	90
AEs / 1000 total patient days	2.3	1.0	2.9	3.0
AEs / 1000 total inpatient days	91.2	104.6	84.3	99.7
AEs / 1000 total outpatient days	0.9	0.5	0.7	1.4
AEs / 100 hospital admissions	79.0	70.8	73.8	89.8
Preventable or mitigable AEs1
No. of AEs identified	115	9	48	58
AEs / 100 patients	29	7	35	43
AEs / 1000 total patient days	0.9	0.0	1.0	1.4
AEs / 1000 total inpatient days	35.3	18.2	31.9	47.6
AEs / 1000 total outpatient days	0.3	0.1	0.2	0.7
AEs / 100 hospital admissions	29.9	13.9	26.2	42.3

¹AEs deemed probably or definitely preventable or mitigable by physician reviewers.

Thirty-four percent of patients had at least one AE (95% CI: 29-39%), and 16% (95% CI: 1320%) of patients had a least one preventable or mitigable AE (Table 3). Approximately half of all AEs occurred within 3 months of the first treatment. Of patients with at least one AE, half had 2 or more AEs over the study period (range 1-9). For patients with multiple events, most AEs (74%) occurred within 30 days of each other.

Table 3.

Patients with adverse events (AEs) over a 1 year time period, by cancer type and stage¹

Cancer type	No. of patients	No. (%) of patients with at least one AE	No. (%) of patients with at least one preventable or mitigable AE
Total	400	136 (34)	64 (16)
Breast cancer	128	27 (21)	9 (7)
Early stage	95	20 (21)	8 (8)
Advanced stage	33	7 (21)	1 (3)
Colorectal cancer	136	56 (41)	26 (19)
Early stage	65	18 (28)	6 (9)
Advanced stage	71	38 (52)	20 (28)
Lung cancer	136	53 (39)	29 (21)
Early stage	68	24 (35)	14 (21)
Advanced stage	68	29 (43)	15 (22)

¹Breast and colorectal cancers: Early, Stages 0-III; Advanced, Stage IV; Lung cancer: Early, NSCLC Stages I-III and SCLC Stage limited; Advanced, SCLC Stage IV and NSCLC extensive

Physician reviewers judged 6% of overall AEs and 4% of preventable AEs to have resulted in permanent harm, required life sustaining intervention, or resulted in death (Harm Categories G-I). For breast cancer, the proportions were 13% and 0%, for colorectal cancer, 4% and 3%, and for lung cancer, 5% and 6%, respectively (Table 4 and Online-only supplemental material 3). Two AEs resulted in death for a rate of 0.01 events per 1000 patient days or 0.52 events per 100 hospital admissions.

Table 4.

Adverse events (AEs) by harm severity¹ and preventability²

Harm Category	No. total AEs (% of total)	No. preventable (% of total) AEs	No. non-preventable (% of total) AEs	No. unable to determine preventability (% of total)
All patients
Total	304	97	147	60
Serious (D-F)	287 (94)	93 (96)	137 (93)	57 (95)
Permanent (G-I)	17 (6)	4 (4)	10 (7)	3 (5)
Breast cancer
Total	46	5	28	13
Serious (D-F)	40 (87)	5 (100)	23 (82)	12 (92)
Permanent (G-I)	6 (13)	0 (0)	5 (18)	1 (8)
Colorectal cancer
Total	135	39	64	32
Serious (D-F)	130 (96)	38 (97)	61 (95)	31 (97)
Permanent (G-I)	5 (4)	1 (3)	3 (5)	1 (3)
Lung cancer
Total	123	53	55	15
Serious (D-F)	117 (95)	50 (94)	53 (96)	14 (93)
Permanent (G-I)	6 (5)	3 (6)	2 (4)	1 (7)

¹Level of harm: D) required monitoring to confirm that it resulted in no harm to the patient and/or required intervention to preclude harm; E) contributed to or resulted in temporary harm to the patient and required intervention; F) contributed to or resulted in temporary harm to patient and required initial or prolonged hospitalization; G) contributed to or resulted in permanent patient harm; H) contributed to or required intervention to sustain life; I) contributed to the patient’s death.

²Preventable AEs refer to those deemed definitely or probably preventable by physician reviewers.

AEs by setting

Sixty-three percent of AEs occurred in the inpatient setting for a rate of 79.0 AEs per 100 hospital admissions and 29.9 preventable or mitigable AEs per 100 hospital admissions. The AE rate was 91.2 per 1000 inpatient days, and 0.9 per 1000 outpatient days. For preventable AEs, the rates were 31.52 per 1000 inpatient days, and 0.24 per 1000 outpatient days. For mitigable AEs, the rates were 3.82 and 0.08, respectively. The AE and preventable or mitigable AE rates by setting differed by cancer type and were mostly the highest for patients with lung cancer (Table 2 and Online-only supplemental material 3).

Differences by cancer type and stage

The overall AE rate among patients with breast cancer was lower than that for patients with colorectal or lung cancers (Kruskal Wallis p≤0.001 for pairwise comparisons) (Table 2). However, a higher proportion of their AEs were permanent (i.e., highest severity) (13% compared to 4% and 5%, respectively) (Table 3). The proportions of patients with an AE or a preventable / mitigable AE were higher among those with advanced stage disease compared to those with early stage disease for patients with colorectal cancer (52% versus 28% for AEs; 28% versus 9% for preventable / mitigable AEs), but not for patients with breast or lung cancers (Table 4).

AE types

The most frequent types of overall AEs identified were those related to infectious events (e.g., Clostridium difficile) (n=50), gastrointestinal system (e.g., mucositis) (n=48), metabolic events (e.g., hypokalemia) (n=47), and hematologic events (e.g., thrombocytopenia) (n=37) (Table 5; see Online-only supplemental material 4 for categorization). Examples of preventable AEs included hypomagnesemia and pressure ulcers. Of the AEs deemed preventable and/or mitigable, the permanent AEs included lymphedema for breast cancer patients, abscess for colorectal cancer patients, and renal failure for lung cancer patients.

Table 5.

Categories of AEs, by cancer type

Adverse Event Category1	Total #	Breast Cancer #	Colorectal Cancer #	Lung Cancer #
Total no. of adverse events	304	46	135	123
Infection	50	9	29	12
Gastrointestinal	48	9	30	9
Metabolic	47	2	12	33
Hematologic	37	3	17	17
Neurologic	24	6	9	9
Genito-urinary	22	0	7	15
Vascular	22	2	12	8
Respiratory	13	1	6	6
Integumentary	11	10	1	0
Cardiovascular	9	0	3	6
Other	8	3	3	2
Medication	7	0	3	4
Delay	3	0	3	0
Fall	3	1	0	2

¹Categories were assigned using clinical judgment. Examples are in Online-only supplemental material 1.

AE data source

Most AEs were identified through medical record review only (85%, 257 AEs). Thirty-three AEs (11%) were identified in both medical records and a patient safety database. Of the overlapping AEs, most (22 AEs) were classified as Harm Category F and occurred during a hospitalization.

DISCUSSION

While clinical oncologists and researchers attend routinely to treatment-related toxicities, relatively little is known about the incidence or nature of errors or treatment-related injuries that affect cancer patients in the course of treatment. In this longitudinal cohort study of 400 patients with breast, colorectal, and lung cancers, patients experienced 304 AEs. In one-third of events, the injury was judged preventable or that the severity or duration of harm could have been reduced. One in three patients had at least one AE in the year in which treatment began. Many AEs were serious and resulted in permanent harm or death. Clearly, cancer patients have a significant burden of care-related harm.

While it is no surprise that cancer patients experience treatment-related toxicities, the extent of treatment-related harm has not been well quantified. Clinical trials frequently exclude “expected” toxicities that fall below a threshold of severity. Trials rarely report out toxicities that are the result of protocol violations or deviations or usual-care lapses as preventable events. There is rarely systematic investigation of AEs affecting patients receiving routine care or “per protocol” treatment outside of a clinical trial or over extended periods. The literature on AEs among cancer patients in routine care is extraordinarily sparse. Most of the work in oncology focuses on medication-specific AEs.^{2, 15}

Few studies assess harm across the continuum of cancer care in a longitudinal way, as this study does. For outpatient events, we found the rates to be 0.9 total AEs and 0.3 preventable or mitigable AEs per 1000 outpatient days. Findings for our inpatient AEs in an oncology population (91.2 AEs and 35.3 preventable or mitigable AEs per 1000 inpatient days) are in line with published studies that used the trigger tool methodology for medical record review to identify AEs in general inpatient populations. Landrigan et. al. conducted an AE assessment in 10 hospitals in North Carolina between 2002 and 2007. Investigators found 25.1 harms per 100 admissions, or 56.5 harms per 1000 patient days. Sixty-three percent of AEs were rated as preventable.²¹ Another study by Classen and colleagues in three hospitals found higher rates: 49 events per 100 admissions, and 91 events per 1000 patient days.²² In the pediatric inpatient population, Kirkendall and colleagues found 36.7 AEs per 100 admissions and 76 AEs per 1000 patient days at a single institution.²³ In our study focused on oncology patients, we found the inpatient AE rate per admission to be higher, at 79.0 AEs per 100 hospital admissions and 29.9 preventable or mitigable AEs per 100 hospital admissions. In line with Classen’s study, we found 91.2 events per 1000 patient days and a preventable or mitigable AE rate of 35.3 events per 1000 inpatient days. While cancer patients are vulnerable to experiencing AEs, these findings suggest that they are in line with rates of adverse events in other studied hospitalized patient populations.^21–23

A single minded focus on cancer treatment toxicities may fail to account for vulnerabilities related to other interventions (e.g., adjunctive therapies, routine medications). While expected complications may arise in surgery or radiation therapy, and toxicities may be related to the properties inherent in cytotoxic or targeted therapies, the use of these therapies may not be optimized and may result in avoidable or prolonged symptoms or side effects. This study contributes to our understanding of the burden of harm in cancer care and offers a methodological contribution by using an oncology-specific screening tool to identify AEs. This extends previous work on AE assessment in general populations to oncology settings. Given recent interest in episode-based reimbursement models in cancer, AEs could be a major cost to providers.^{24, 25} In sum, our study identifies opportunities for oncology clinicians to think more broadly about oncology-related harm and the interventions that may identify, intercept, prevent, or mitigate injuries.^26–28

There are some limitations. First, the study was performed at a single institution with a broad referral population and strong clinical trials program. The patient population is more likely to reflect care at other regional referral centers than community sites. The institution’s emphasis on patient safety may both increase detection and reporting of AEs but also reduce the incidence compared to other centers. Second, we relied on medical record review and two voluntary reporting systems for AE identification. Most of the existing reports were for close-calls. We likely underestimated the number of AEs, because medical record documentation is often incomplete.²⁹ However, these limitations are inherent in population-based medical record studies of AEs and medical errors. Also, some AEs that occurred in the outpatient setting might be incorrectly classified as inpatient, and vice versa, given the unknown definitive cause of the AE. Finally, the likelihood of an AE’s preventability or harm mitigation was a subjective assessment by expert physician reviewers. This assessment is inherently challenging in cancer care given the difficulty distinguishing expected disease-related toxicities from unnecessary harm. However, we required two reviewers to consider the context and current best practice and agree on each case, and had generally good agreement.

Systematically generated information about AEs that patients experience over the course of cancer care can have several purposes. It can be used to identify issues that are common and harmful and direct interventions to prevent or mitigate the harm associated with these events. Patients and providers can be informed about what harms they might anticipate during the course of care. This information can also aid in the development of cancer-specific quality measures that can facilitate monitoring and assessing specific AEs on a regular basis.³⁰ In future iterations, oncology signals can be automated to facilitate identification of high-risk AEs and potential opportunities to intervene in real-time.

Our results provide insights into the nature and extent of harm, including preventable and mitigable harm, experienced by patients during cancer care over one year at a single comprehensive cancer center. Given ongoing developments in cancer therapies, clinicians and systems need to examine the safety of their own practices and develop approaches to address potentially preventable harm.